From bf74c581ea60798ff46b23c33370aedbcaa80593 Mon Sep 17 00:00:00 2001
From: Frank Steiler <frank@steiler.dev>
Date: Mon, 3 Aug 2026 13:06:03 +0200
Subject: [PATCH 1/6] =?UTF-8?q?fix(reports):=20improve=20report=20PDF=20UX?=
 =?UTF-8?q?=20=E2=80=94=20paragraph=20breaks,=20inline=20meta,=20inline=20?=
 =?UTF-8?q?notes,=20column=20toggles?=
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

- AI cover letter body now splits on double newlines into separate pdfmake
  content blocks, preserving paragraph breaks from the LLM response
- Area and attachments count moved inline into the usage cell (grey, same
  font size) instead of a separate stack/sub-row
- Split (†) and deposit-reduced (‡) footnotes replaced with inline grey
  labels in the allocated amount cell, matching the existing deposit pattern;
  allocatedMarkers field removed from ReportContentRow in favour of boolean
  isSplit / isDepositReduced flags
- ReportContentEditor gains per-column visibility toggles (local state, no
  persistence); attachments column removed — content is now inline under usage

Co-Authored-By: Claude frontend-developer <noreply@anthropic.com>
---
 .../reports/ReportContentEditor.module.css    |  36 ++-
 .../reports/ReportContentEditor.test.tsx      |  16 +-
 .../reports/ReportContentEditor.tsx           | 300 ++++++++++--------
 client/src/i18n/de/budget.json                |   5 +-
 client/src/i18n/en/budget.json                |   5 +-
 .../lib/reportContent/applyAiContent.test.ts  |   5 +-
 .../lib/reportContent/applyOverrides.test.ts  |   5 +-
 .../reportContent/buildReportContent.test.ts  | 124 +++-----
 .../lib/reportContent/buildReportContent.ts   |  33 +-
 client/src/lib/reportContent/types.ts         |   7 +-
 client/src/lib/reportPdf/coverLetterPdf.ts    |  23 +-
 client/src/lib/reportPdf/overviewPdf.test.ts  |  97 +++---
 client/src/lib/reportPdf/overviewPdf.ts       |  92 ++----
 client/src/lib/reportPdf/realRender.test.ts   |   8 +-
 14 files changed, 384 insertions(+), 372 deletions(-)

diff --git a/client/src/components/reports/ReportContentEditor.module.css b/client/src/components/reports/ReportContentEditor.module.css
index a88bde78e..ba129259f 100644
--- a/client/src/components/reports/ReportContentEditor.module.css
+++ b/client/src/components/reports/ReportContentEditor.module.css
@@ -67,12 +67,37 @@
   gap: var(--spacing-1);
 }
 
-/* Table Heading */
+/* Table Heading + Column Toggles */
+.tableHeadingRow {
+  display: flex;
+  align-items: flex-start;
+  gap: var(--spacing-4);
+  flex-wrap: wrap;
+}
+
 .tableHeading {
   margin: 0 0 var(--spacing-3) 0;
   font-size: var(--font-size-lg);
   font-weight: var(--font-weight-semibold);
   color: var(--color-text-primary);
+  flex-shrink: 0;
+}
+
+.columnToggles {
+  display: flex;
+  flex-wrap: wrap;
+  gap: var(--spacing-2);
+  margin-bottom: var(--spacing-3);
+}
+
+.columnToggle {
+  display: flex;
+  align-items: center;
+  gap: var(--spacing-1);
+  font-size: var(--font-size-xs);
+  color: var(--color-text-muted);
+  cursor: pointer;
+  white-space: nowrap;
 }
 
 /* Table Wrapper (Desktop) */
@@ -170,12 +195,17 @@
   flex-wrap: wrap;
 }
 
-.usageAreaText {
-  font-size: var(--font-size-xs);
+.usageMetaText {
   color: var(--color-text-muted);
   margin-top: var(--spacing-1);
 }
 
+.inlineNote {
+  font-size: var(--font-size-xs);
+  color: var(--color-text-muted);
+  margin-left: var(--spacing-1);
+}
+
 /* Summary Table */
 .summaryTable {
   width: 100%;
diff --git a/client/src/components/reports/ReportContentEditor.test.tsx b/client/src/components/reports/ReportContentEditor.test.tsx
index 1e2f6b585..042af6fa2 100644
--- a/client/src/components/reports/ReportContentEditor.test.tsx
+++ b/client/src/components/reports/ReportContentEditor.test.tsx
@@ -79,6 +79,8 @@ const LABELS: ReportContentLabels = {
   usage: 'REPORT_USAGE_LABEL',
   attachmentsNote: 'REPORT_ATTACHMENTS_NOTE_LABEL',
   deposit: 'REPORT_DEPOSIT_LABEL',
+  splitNote: 'REPORT_SPLIT_NOTE_LABEL',
+  depositReducedNote: 'REPORT_DEPOSIT_REDUCED_NOTE_LABEL',
   source: 'REPORT_SOURCE_LABEL',
   sourceType: 'REPORT_SOURCE_TYPE_LABEL',
   reference: 'REPORT_REFERENCE_LABEL',
@@ -95,7 +97,8 @@ function makeRow(overrides: Partial<ReportContentRow> = {}): ReportContentRow {
     statusText: null,
     invoiceAmountText: '€100.00',
     allocatedAmountValueText: '€100.00',
-    allocatedMarkers: '',
+    isSplit: false,
+    isDepositReduced: false,
     isDeposit: false,
     isRefund: false,
     refundNoteText: '',
@@ -704,18 +707,17 @@ describe('ReportContentEditor — table rows', () => {
     expect(within(table).queryByDisplayValue('€555.00')).not.toBeInTheDocument();
   });
 
-  it('composes the allocated cell as valueText + markers + refund note when isRefund', () => {
+  it('composes the allocated cell as valueText + refund note when isRefund', () => {
     const rows = [
       makeRow({
         allocatedAmountValueText: '€-200.00',
-        allocatedMarkers: '†1',
         isRefund: true,
         refundNoteText: '(refund)',
       }),
     ];
     const { container } = renderEditor({ content: makeContent({ rows }) });
     const table = getDesktopTable(container);
-    expect(within(table).getByText('€-200.00†1 (refund)')).toBeInTheDocument();
+    expect(within(table).getByText('€-200.00 (refund)')).toBeInTheDocument();
   });
 
   it('applies the refundAmount CSS class (not an inline style) to both amount cells when isRefund', () => {
@@ -913,7 +915,6 @@ describe('ReportContentEditor — isDeposit (AC2.1: inline Deposit badge, no mar
       makeRow({
         invoiceId: 'inv-1',
         isDeposit: true,
-        allocatedMarkers: '',
         allocatedAmountValueText: '€300.00',
       }),
     ];
@@ -1054,18 +1055,17 @@ describe(
       expect(card.queryByLabelText(LABELS.attachmentsNote)).not.toBeInTheDocument();
     });
 
-    it('composes the mobile card allocated amount as valueText + markers + refund note when isRefund, matching the desktop cell', () => {
+    it('composes the mobile card allocated amount as valueText + refund note when isRefund, matching the desktop cell', () => {
       const rows = [
         makeRow({
           allocatedAmountValueText: '€-200.00',
-          allocatedMarkers: '†1',
           isRefund: true,
           refundNoteText: '(refund)',
         }),
       ];
       const { container } = renderEditor({ content: makeContent({ rows }) });
       const card = within(getMobileList(container));
-      expect(card.getByText('€-200.00†1 (refund)')).toBeInTheDocument();
+      expect(card.getByText('€-200.00 (refund)')).toBeInTheDocument();
     });
 
     it('wires the mobile card usage/attachmentsNote EditableFields to the same onFieldChange keys as the desktop table', () => {
diff --git a/client/src/components/reports/ReportContentEditor.tsx b/client/src/components/reports/ReportContentEditor.tsx
index c81a2b98e..d019bc87f 100644
--- a/client/src/components/reports/ReportContentEditor.tsx
+++ b/client/src/components/reports/ReportContentEditor.tsx
@@ -3,6 +3,7 @@
  * Handles field changes and resets via callbacks; no state management.
  */
 
+import { useState } from 'react';
 import type { TFunction } from 'i18next';
 import type { InvoiceStatus } from '@cornerstone/shared';
 import type { ReportContent, ReportContentOverrides } from '../../lib/reportContent/index.js';
@@ -27,6 +28,9 @@ const STATUS_BADGE_CLASSNAME: Record<InvoiceStatus, string> = {
   quotation: styles.statusQuotation!,
 };
 
+type ColumnKey =
+  'vendor' | 'invoiceNumber' | 'date' | 'status' | 'invoiceAmount' | 'allocatedAmount' | 'usage';
+
 export function ReportContentEditor({
   content,
   overrides,
@@ -37,6 +41,18 @@ export function ReportContentEditor({
   // Helper: check if a field has been overridden
   const isFieldEdited = (key: string): boolean => key in overrides;
 
+  // Column visibility state
+  const [hiddenColumns, setHiddenColumns] = useState<Set<ColumnKey>>(new Set());
+  const toggleColumn = (col: ColumnKey) => {
+    setHiddenColumns((prev) => {
+      const next = new Set(prev);
+      if (next.has(col)) next.delete(col);
+      else next.add(col);
+      return next;
+    });
+  };
+  const show = (col: ColumnKey) => !hiddenColumns.has(col);
+
   return (
     <div className={styles.container}>
       {/* Cover Letter */}
@@ -174,30 +190,57 @@ export function ReportContentEditor({
       )}
 
       {/* Report Table */}
-      <h3 className={styles.tableHeading}>{t('sourceReports.editable.tableHeading')}</h3>
+      <div className={styles.tableHeadingRow}>
+        <h3 className={styles.tableHeading}>{t('sourceReports.editable.tableHeading')}</h3>
+        <div
+          className={styles.columnToggles}
+          role="group"
+          aria-label={t('sourceReports.editable.columnVisibilityLabel')}
+        >
+          {(
+            [
+              ['vendor', content.labels.vendor],
+              ['invoiceNumber', content.labels.invoiceNumber],
+              ['date', content.labels.date],
+              ...(content.isOverview
+                ? [['status', content.labels.status] as [ColumnKey, string]]
+                : []),
+              ['invoiceAmount', content.labels.invoiceAmount],
+              ['allocatedAmount', content.labels.allocatedAmount],
+              ['usage', content.labels.usage],
+            ] as [ColumnKey, string][]
+          ).map(([col, label]) => (
+            <label key={col} className={styles.columnToggle}>
+              <input type="checkbox" checked={show(col)} onChange={() => toggleColumn(col)} />
+              {label}
+            </label>
+          ))}
+        </div>
+      </div>
       <div className={styles.tableWrapper}>
         <table className={styles.table}>
           <thead>
             <tr>
-              <th>{content.labels.vendor}</th>
-              <th>{content.labels.invoiceNumber}</th>
-              <th>{content.labels.date}</th>
-              {content.isOverview && <th>{content.labels.status}</th>}
-              <th className={styles.rightAlign}>{content.labels.invoiceAmount}</th>
-              <th className={styles.rightAlign}>{content.labels.allocatedAmount}</th>
-              <th>{content.labels.usage}</th>
-              {content.rows.some((r) => r.attachmentsNote !== null) && (
-                <th>{content.labels.attachmentsNote}</th>
+              {show('vendor') && <th>{content.labels.vendor}</th>}
+              {show('invoiceNumber') && <th>{content.labels.invoiceNumber}</th>}
+              {show('date') && <th>{content.labels.date}</th>}
+              {content.isOverview && show('status') && <th>{content.labels.status}</th>}
+              {show('invoiceAmount') && (
+                <th className={styles.rightAlign}>{content.labels.invoiceAmount}</th>
+              )}
+              {show('allocatedAmount') && (
+                <th className={styles.rightAlign}>{content.labels.allocatedAmount}</th>
               )}
+              {show('usage') && <th>{content.labels.usage}</th>}
             </tr>
           </thead>
           <tbody>
             {content.rows.map((row) => (
               <tr key={row.invoiceId}>
-                <td>{row.vendor}</td>
-                <td>{row.invoiceNumber}</td>
-                <td>{row.dateText}</td>
-                {content.isOverview && row.status && row.statusText != null && (
+                {show('vendor') && <td>{row.vendor}</td>}
+                {show('invoiceNumber') && <td>{row.invoiceNumber}</td>}
+                {show('date') && <td>{row.dateText}</td>}
+                {content.isOverview && show('status') && row.status && row.statusText != null && (
                   <td>
                     <Badge
                       value={row.status}
@@ -210,65 +253,65 @@ export function ReportContentEditor({
                     />
                   </td>
                 )}
-                <td className={`${styles.rightAlign} ${row.isRefund ? styles.refundAmount : ''}`}>
-                  {row.invoiceAmountText}
-                </td>
-                <td className={`${styles.rightAlign} ${row.isRefund ? styles.refundAmount : ''}`}>
-                  {row.allocatedAmountValueText}
-                  {row.allocatedMarkers}
-                  {row.isRefund && ` ${row.refundNoteText}`}
-                  {row.isDeposit && (
-                    <Badge
-                      className={styles.depositLabel}
-                      variants={{
-                        deposit: {
-                          label: content.labels.deposit,
-                          className: styles.depositBadge,
-                        },
-                      }}
-                      value="deposit"
-                    />
-                  )}
-                </td>
-                <td>
-                  <EditableField
-                    as="input"
-                    ariaLabel={t('sourceReports.editable.usageTextAriaLabel', {
-                      vendor: row.vendor,
-                      invoiceNumber: row.invoiceNumber,
-                    })}
-                    editedSuffix={t('sourceReports.editable.editedSuffix')}
-                    resetAriaLabel={t('sourceReports.editable.resetFieldAriaLabel', {
-                      field: t('sourceReports.table.usage'),
-                    })}
-                    value={row.usageText}
-                    onChange={(value) =>
-                      onFieldChange(overrideKey.row(row.invoiceId).usageText, value)
-                    }
-                    isEdited={isFieldEdited(overrideKey.row(row.invoiceId).usageText)}
-                    onReset={() => onFieldReset(overrideKey.row(row.invoiceId).usageText)}
-                  />
-                  {row.areaText && <div className={styles.usageAreaText}>{row.areaText}</div>}
-                </td>
-                {row.attachmentsNote !== null && (
+                {content.isOverview &&
+                  show('status') &&
+                  (!row.status || row.statusText == null) && <td />}
+                {show('invoiceAmount') && (
+                  <td className={`${styles.rightAlign} ${row.isRefund ? styles.refundAmount : ''}`}>
+                    {row.invoiceAmountText}
+                  </td>
+                )}
+                {show('allocatedAmount') && (
+                  <td className={`${styles.rightAlign} ${row.isRefund ? styles.refundAmount : ''}`}>
+                    {row.allocatedAmountValueText}
+                    {row.isRefund && ` ${row.refundNoteText}`}
+                    {row.isDeposit && (
+                      <Badge
+                        className={styles.depositLabel}
+                        variants={{
+                          deposit: {
+                            label: content.labels.deposit,
+                            className: styles.depositBadge,
+                          },
+                        }}
+                        value="deposit"
+                      />
+                    )}
+                    {row.isSplit && (
+                      <span className={styles.inlineNote}> ({content.labels.splitNote})</span>
+                    )}
+                    {row.isDepositReduced && (
+                      <span className={styles.inlineNote}>
+                        {' '}
+                        ({content.labels.depositReducedNote})
+                      </span>
+                    )}
+                  </td>
+                )}
+                {show('usage') && (
                   <td>
                     <EditableField
                       as="input"
-                      ariaLabel={t('sourceReports.editable.attachmentsNoteAriaLabel', {
+                      ariaLabel={t('sourceReports.editable.usageTextAriaLabel', {
                         vendor: row.vendor,
                         invoiceNumber: row.invoiceNumber,
                       })}
                       editedSuffix={t('sourceReports.editable.editedSuffix')}
                       resetAriaLabel={t('sourceReports.editable.resetFieldAriaLabel', {
-                        field: t('sourceReports.editable.attachmentsNoteLabel'),
+                        field: t('sourceReports.table.usage'),
                       })}
-                      value={row.attachmentsNote}
+                      value={row.usageText}
                       onChange={(value) =>
-                        onFieldChange(overrideKey.row(row.invoiceId).attachmentsNote, value)
+                        onFieldChange(overrideKey.row(row.invoiceId).usageText, value)
                       }
-                      isEdited={isFieldEdited(overrideKey.row(row.invoiceId).attachmentsNote)}
-                      onReset={() => onFieldReset(overrideKey.row(row.invoiceId).attachmentsNote)}
+                      isEdited={isFieldEdited(overrideKey.row(row.invoiceId).usageText)}
+                      onReset={() => onFieldReset(overrideKey.row(row.invoiceId).usageText)}
                     />
+                    {(row.areaText || row.attachmentsNote) && (
+                      <div className={styles.usageMetaText}>
+                        {[row.areaText, row.attachmentsNote].filter(Boolean).join(' · ')}
+                      </div>
+                    )}
                   </td>
                 )}
               </tr>
@@ -281,19 +324,25 @@ export function ReportContentEditor({
       <div className={styles.mobileCardList}>
         {content.rows.map((row) => (
           <div key={row.invoiceId} className={styles.mobileCard}>
-            <div className={styles.mobileCardRow}>
-              <span className={styles.mobileCardCaption}>{content.labels.vendor}</span>
-              <span className={styles.mobileCardValue}>{row.vendor}</span>
-            </div>
-            <div className={styles.mobileCardRow}>
-              <span className={styles.mobileCardCaption}>{content.labels.invoiceNumber}</span>
-              <span className={styles.mobileCardValue}>{row.invoiceNumber}</span>
-            </div>
-            <div className={styles.mobileCardRow}>
-              <span className={styles.mobileCardCaption}>{content.labels.date}</span>
-              <span className={styles.mobileCardValue}>{row.dateText}</span>
-            </div>
-            {content.isOverview && row.status && row.statusText != null && (
+            {show('vendor') && (
+              <div className={styles.mobileCardRow}>
+                <span className={styles.mobileCardCaption}>{content.labels.vendor}</span>
+                <span className={styles.mobileCardValue}>{row.vendor}</span>
+              </div>
+            )}
+            {show('invoiceNumber') && (
+              <div className={styles.mobileCardRow}>
+                <span className={styles.mobileCardCaption}>{content.labels.invoiceNumber}</span>
+                <span className={styles.mobileCardValue}>{row.invoiceNumber}</span>
+              </div>
+            )}
+            {show('date') && (
+              <div className={styles.mobileCardRow}>
+                <span className={styles.mobileCardCaption}>{content.labels.date}</span>
+                <span className={styles.mobileCardValue}>{row.dateText}</span>
+              </div>
+            )}
+            {content.isOverview && show('status') && row.status && row.statusText != null && (
               <div className={styles.mobileCardRow}>
                 <span className={styles.mobileCardCaption}>{content.labels.status}</span>
                 <Badge
@@ -307,76 +356,71 @@ export function ReportContentEditor({
                 />
               </div>
             )}
-            <div className={styles.mobileCardRow}>
-              <span className={styles.mobileCardCaption}>{content.labels.invoiceAmount}</span>
-              <span
-                className={`${styles.mobileCardValue} ${row.isRefund ? styles.refundAmount : ''}`}
-              >
-                {row.invoiceAmountText}
-              </span>
-            </div>
-            <div className={styles.mobileCardRow}>
-              <span className={styles.mobileCardCaption}>{content.labels.allocatedAmount}</span>
-              <span className={styles.mobileCardAllocated}>
+            {show('invoiceAmount') && (
+              <div className={styles.mobileCardRow}>
+                <span className={styles.mobileCardCaption}>{content.labels.invoiceAmount}</span>
                 <span
                   className={`${styles.mobileCardValue} ${row.isRefund ? styles.refundAmount : ''}`}
                 >
-                  {row.allocatedAmountValueText}
-                  {row.allocatedMarkers}
-                  {row.isRefund && ` ${row.refundNoteText}`}
+                  {row.invoiceAmountText}
                 </span>
-                {row.isDeposit && (
-                  <Badge
-                    variants={{
-                      deposit: {
-                        label: content.labels.deposit,
-                        className: styles.depositBadge,
-                      },
-                    }}
-                    value="deposit"
-                  />
-                )}
-              </span>
-            </div>
-            <div className={styles.mobileCardRow}>
-              <EditableField
-                as="input"
-                label={content.labels.usage}
-                ariaLabel={t('sourceReports.editable.usageTextAriaLabel', {
-                  vendor: row.vendor,
-                  invoiceNumber: row.invoiceNumber,
-                })}
-                editedSuffix={t('sourceReports.editable.editedSuffix')}
-                resetAriaLabel={t('sourceReports.editable.resetFieldAriaLabel', {
-                  field: t('sourceReports.table.usage'),
-                })}
-                value={row.usageText}
-                onChange={(value) => onFieldChange(overrideKey.row(row.invoiceId).usageText, value)}
-                isEdited={isFieldEdited(overrideKey.row(row.invoiceId).usageText)}
-                onReset={() => onFieldReset(overrideKey.row(row.invoiceId).usageText)}
-              />
-              {row.areaText && <span className={styles.usageAreaText}>{row.areaText}</span>}
-            </div>
-            {row.attachmentsNote !== null && (
+              </div>
+            )}
+            {show('allocatedAmount') && (
+              <div className={styles.mobileCardRow}>
+                <span className={styles.mobileCardCaption}>{content.labels.allocatedAmount}</span>
+                <span className={styles.mobileCardAllocated}>
+                  <span
+                    className={`${styles.mobileCardValue} ${row.isRefund ? styles.refundAmount : ''}`}
+                  >
+                    {row.allocatedAmountValueText}
+                    {row.isRefund && ` ${row.refundNoteText}`}
+                  </span>
+                  {row.isDeposit && (
+                    <Badge
+                      variants={{
+                        deposit: {
+                          label: content.labels.deposit,
+                          className: styles.depositBadge,
+                        },
+                      }}
+                      value="deposit"
+                    />
+                  )}
+                  {row.isSplit && (
+                    <span className={styles.inlineNote}>({content.labels.splitNote})</span>
+                  )}
+                  {row.isDepositReduced && (
+                    <span className={styles.inlineNote}>({content.labels.depositReducedNote})</span>
+                  )}
+                </span>
+              </div>
+            )}
+            {show('usage') && (
               <div className={styles.mobileCardRow}>
                 <EditableField
                   as="input"
-                  label={content.labels.attachmentsNote}
-                  ariaLabel={t('sourceReports.editable.attachmentsNoteAriaLabel', {
+                  label={content.labels.usage}
+                  ariaLabel={t('sourceReports.editable.usageTextAriaLabel', {
                     vendor: row.vendor,
                     invoiceNumber: row.invoiceNumber,
                   })}
                   editedSuffix={t('sourceReports.editable.editedSuffix')}
                   resetAriaLabel={t('sourceReports.editable.resetFieldAriaLabel', {
-                    field: t('sourceReports.editable.attachmentsNoteLabel'),
+                    field: t('sourceReports.table.usage'),
                   })}
-                  value={row.attachmentsNote}
+                  value={row.usageText}
                   onChange={(value) =>
-                    onFieldChange(overrideKey.row(row.invoiceId).attachmentsNote, value)
+                    onFieldChange(overrideKey.row(row.invoiceId).usageText, value)
                   }
-                  isEdited={isFieldEdited(overrideKey.row(row.invoiceId).attachmentsNote)}
-                  onReset={() => onFieldReset(overrideKey.row(row.invoiceId).attachmentsNote)}
+                  isEdited={isFieldEdited(overrideKey.row(row.invoiceId).usageText)}
+                  onReset={() => onFieldReset(overrideKey.row(row.invoiceId).usageText)}
                 />
+                {(row.areaText || row.attachmentsNote) && (
+                  <span className={styles.usageMetaText}>
+                    {[row.areaText, row.attachmentsNote].filter(Boolean).join(' · ')}
+                  </span>
+                )}
               </div>
             )}
           </div>
diff --git a/client/src/i18n/de/budget.json b/client/src/i18n/de/budget.json
index 95b65fa31..572cf2640 100644
--- a/client/src/i18n/de/budget.json
+++ b/client/src/i18n/de/budget.json
@@ -1215,6 +1215,7 @@
       "keepEditing": "Weiter bearbeiten",
       "coverLetterHeading": "Anschreiben",
       "tableHeading": "Berichtstabelle",
+      "columnVisibilityLabel": "Spalten ein-/ausblenden",
       "enhanceWithAi": "Mit KI verbessern",
       "enhanceWithAiDescription": "Ersetzt die unten stehenden Verwendungstexte und das Anschreiben durch KI-generierte Inhalte. Vorgenommene Bearbeitungen gehen dabei verloren.",
       "generating": "Generiere… ({{seconds}}s)",
@@ -1256,7 +1257,9 @@
       "attachmentsNote_other": "{{count}} Anhänge: {{types}}",
       "attachmentsNoteNoType_one": "{{count}} Anhang",
       "attachmentsNoteNoType_other": "{{count}} Anhänge",
-      "depositReducedFootnote": "Diese Position berücksichtigt separat eingereichte Abschlagszahlungen."
+      "depositReducedFootnote": "Diese Position berücksichtigt separat eingereichte Abschlagszahlungen.",
+      "splitInlineLabel": "Teilbetrag",
+      "depositReducedInlineLabel": "abzgl. Abschlag"
     },
     "sourceType": {
       "bank_loan": "Bankdarlehen",
diff --git a/client/src/i18n/en/budget.json b/client/src/i18n/en/budget.json
index 882a2f6f5..1921c8803 100644
--- a/client/src/i18n/en/budget.json
+++ b/client/src/i18n/en/budget.json
@@ -1215,6 +1215,7 @@
       "keepEditing": "Keep Editing",
       "coverLetterHeading": "Cover Letter",
       "tableHeading": "Report Table",
+      "columnVisibilityLabel": "Show/hide columns",
       "enhanceWithAi": "Enhance with AI",
       "enhanceWithAiDescription": "Replaces the usage descriptions and cover letter below with AI-generated content. Any edits you've made will be discarded.",
       "generating": "Generating… ({{seconds}}s)",
@@ -1256,7 +1257,9 @@
       "footnoteFetchFailed": "Document could not be retrieved",
       "footnoteInvalidPdf": "Document is not a valid PDF",
       "splitFootnote": "Amount shown reflects only the portion allocated to this source.",
-      "depositReducedFootnote": "This position reflects deposits claimed separately."
+      "depositReducedFootnote": "This position reflects deposits claimed separately.",
+      "splitInlineLabel": "partial",
+      "depositReducedInlineLabel": "less deposit"
     },
     "sourceType": {
       "bank_loan": "Bank Loan",
diff --git a/client/src/lib/reportContent/applyAiContent.test.ts b/client/src/lib/reportContent/applyAiContent.test.ts
index a8b44d55f..715382c22 100644
--- a/client/src/lib/reportContent/applyAiContent.test.ts
+++ b/client/src/lib/reportContent/applyAiContent.test.ts
@@ -22,7 +22,8 @@ function makeRow(overrides: Partial<ReportContentRow> = {}): ReportContentRow {
     statusText: null,
     invoiceAmountText: '€100.00',
     allocatedAmountValueText: '€100.00',
-    allocatedMarkers: '',
+    isSplit: false,
+    isDepositReduced: false,
     isDeposit: false,
     isRefund: false,
     refundNoteText: '',
@@ -44,6 +45,8 @@ function makeLabels(): ReportContent['labels'] {
     usage: 'Usage',
     attachmentsNote: 'Attachments Note',
     deposit: 'Deposit',
+    splitNote: 'partial',
+    depositReducedNote: 'less deposit',
     source: 'Source',
     sourceType: 'Source Type',
     reference: 'Reference',
diff --git a/client/src/lib/reportContent/applyOverrides.test.ts b/client/src/lib/reportContent/applyOverrides.test.ts
index c92dbf528..1ced2be53 100644
--- a/client/src/lib/reportContent/applyOverrides.test.ts
+++ b/client/src/lib/reportContent/applyOverrides.test.ts
@@ -21,7 +21,8 @@ function makeRow(overrides: Partial<ReportContentRow> = {}): ReportContentRow {
     statusText: null,
     invoiceAmountText: '€100.00',
     allocatedAmountValueText: '€100.00',
-    allocatedMarkers: '',
+    isSplit: false,
+    isDepositReduced: false,
     isDeposit: false,
     isRefund: false,
     refundNoteText: '',
@@ -43,6 +44,8 @@ function makeLabels(): ReportContent['labels'] {
     usage: 'Usage',
     attachmentsNote: 'Attachments Note',
     deposit: 'Deposit',
+    splitNote: 'partial',
+    depositReducedNote: 'less deposit',
     source: 'Source',
     sourceType: 'Source Type',
     reference: 'Reference',
diff --git a/client/src/lib/reportContent/buildReportContent.test.ts b/client/src/lib/reportContent/buildReportContent.test.ts
index 121de3181..75e7a4eb9 100644
--- a/client/src/lib/reportContent/buildReportContent.test.ts
+++ b/client/src/lib/reportContent/buildReportContent.test.ts
@@ -342,8 +342,8 @@ describe('buildReportContent — rows', () => {
     });
   });
 
-  describe('allocatedMarkers (split † / deposit ‡, unnumbered/shared per story #1923) + isDeposit', () => {
-    it('adds unnumbered † only when isSplit and budgetLines.length > 0, no deposits', () => {
+  describe('isSplit / isDepositReduced / isDeposit flags', () => {
+    it('sets isSplit=true when isSplit and budgetLines.length > 0, no deposits', () => {
       const invoice = makeInvoice({
         isSplit: true,
         budgetLines: [makeBudgetLine()],
@@ -351,11 +351,12 @@ describe('buildReportContent — rows', () => {
       });
       const report = makeReport([invoice]);
       const content = buildReportContent(report, new Set(['inv-1']), 'claim', t, formatters);
-      expect(content.rows[0]!.allocatedMarkers).toBe('†');
+      expect(content.rows[0]!.isSplit).toBe(true);
+      expect(content.rows[0]!.isDepositReduced).toBe(false);
       expect(content.rows[0]!.isDeposit).toBe(false);
     });
 
-    it('adds unnumbered ‡ only when isSplit and the deposit is untagged (reduced), no budget lines', () => {
+    it('sets isDepositReduced=true when isSplit and the deposit is untagged (reduced), no budget lines', () => {
       const invoice = makeInvoice({
         isSplit: true,
         budgetLines: [],
@@ -363,11 +364,12 @@ describe('buildReportContent — rows', () => {
       });
       const report = makeReport([invoice], { id: 'src-1' });
       const content = buildReportContent(report, new Set(['inv-1']), 'claim', t, formatters);
-      expect(content.rows[0]!.allocatedMarkers).toBe('‡');
+      expect(content.rows[0]!.isSplit).toBe(false);
+      expect(content.rows[0]!.isDepositReduced).toBe(true);
       expect(content.rows[0]!.isDeposit).toBe(false);
     });
 
-    it('AC2.1: adds NO marker and sets isDeposit=true when the deposit is tagged to this report source (constituted)', () => {
+    it('AC2.1: sets isDeposit=true (not isSplit/isDepositReduced) when the deposit is tagged to this report source (constituted)', () => {
       const invoice = makeInvoice({
         isSplit: true,
         budgetLines: [],
@@ -375,11 +377,12 @@ describe('buildReportContent — rows', () => {
       });
       const report = makeReport([invoice], { id: 'src-1' });
       const content = buildReportContent(report, new Set(['inv-1']), 'claim', t, formatters);
-      expect(content.rows[0]!.allocatedMarkers).toBe('');
+      expect(content.rows[0]!.isSplit).toBe(false);
+      expect(content.rows[0]!.isDepositReduced).toBe(false);
       expect(content.rows[0]!.isDeposit).toBe(true);
     });
 
-    it('AC2.4: adds both † and ‡ (order †‡) when split budget lines and a reduced (untagged) deposit are both present', () => {
+    it('AC2.4: sets both isSplit and isDepositReduced when split budget lines and a reduced deposit are both present', () => {
       const invoice = makeInvoice({
         isSplit: true,
         budgetLines: [makeBudgetLine()],
@@ -387,10 +390,11 @@ describe('buildReportContent — rows', () => {
       });
       const report = makeReport([invoice], { id: 'src-1' });
       const content = buildReportContent(report, new Set(['inv-1']), 'claim', t, formatters);
-      expect(content.rows[0]!.allocatedMarkers).toBe('†‡');
+      expect(content.rows[0]!.isSplit).toBe(true);
+      expect(content.rows[0]!.isDepositReduced).toBe(true);
     });
 
-    it('adds only † (no ‡, isDeposit=true) when split budget lines are combined with a constituted (tagged) deposit', () => {
+    it('sets isSplit=true and isDeposit=true when split budget lines combined with a constituted (tagged) deposit', () => {
       const invoice = makeInvoice({
         isSplit: true,
         budgetLines: [makeBudgetLine()],
@@ -398,11 +402,12 @@ describe('buildReportContent — rows', () => {
       });
       const report = makeReport([invoice], { id: 'src-1' });
       const content = buildReportContent(report, new Set(['inv-1']), 'claim', t, formatters);
-      expect(content.rows[0]!.allocatedMarkers).toBe('†');
+      expect(content.rows[0]!.isSplit).toBe(true);
+      expect(content.rows[0]!.isDepositReduced).toBe(false);
       expect(content.rows[0]!.isDeposit).toBe(true);
     });
 
-    it('adds neither marker nor isDeposit when isSplit is false, regardless of budgetLines/deposits content', () => {
+    it('all flags are false when isSplit is false, regardless of budgetLines/deposits content', () => {
       const invoice = makeInvoice({
         isSplit: false,
         budgetLines: [makeBudgetLine()],
@@ -410,19 +415,21 @@ describe('buildReportContent — rows', () => {
       });
       const report = makeReport([invoice]);
       const content = buildReportContent(report, new Set(['inv-1']), 'claim', t, formatters);
-      expect(content.rows[0]!.allocatedMarkers).toBe('');
+      expect(content.rows[0]!.isSplit).toBe(false);
+      expect(content.rows[0]!.isDepositReduced).toBe(false);
       expect(content.rows[0]!.isDeposit).toBe(false);
     });
 
-    it('adds neither marker when isSplit is true but budgetLines and deposits are both empty', () => {
+    it('all flags are false when isSplit is true but budgetLines and deposits are both empty', () => {
       const invoice = makeInvoice({ isSplit: true, budgetLines: [], deposits: [] });
       const report = makeReport([invoice]);
       const content = buildReportContent(report, new Set(['inv-1']), 'claim', t, formatters);
-      expect(content.rows[0]!.allocatedMarkers).toBe('');
+      expect(content.rows[0]!.isSplit).toBe(false);
+      expect(content.rows[0]!.isDepositReduced).toBe(false);
       expect(content.rows[0]!.isDeposit).toBe(false);
     });
 
-    it('never assigns markers to an excluded invoice, even when isSplit with lines/deposits', () => {
+    it('never sets flags on an excluded invoice, even when isSplit with lines/deposits', () => {
       const invoice = makeInvoice({
         invoiceId: 'inv-excluded',
         isSplit: true,
@@ -431,63 +438,38 @@ describe('buildReportContent — rows', () => {
       const included = makeInvoice({ invoiceId: 'inv-1' });
       const report = makeReport([invoice, included]);
       const content = buildReportContent(report, new Set(['inv-1']), 'claim', t, formatters);
-      // Only the included row is present; footnotes must not have been generated for the excluded one.
       expect(content.rows).toHaveLength(1);
       expect(content.footnotes).toEqual([]);
     });
   });
 });
 
-describe('buildReportContent — footnotes (AC1/AC2: shared, unnumbered, at most 2 entries, no vendor prefix)', () => {
-  it('AC1.2: produces exactly one shared split footnote (no vendor/invoice-number prefix) when three included invoices are split', () => {
+describe('buildReportContent — footnotes (always empty; split/deposit annotations are inline)', () => {
+  it('produces no footnotes when invoices are split with budget lines', () => {
     const inv1 = makeInvoice({
       invoiceId: 'inv-1',
-      vendorName: 'Gamma Corp',
-      invoiceNumber: 'G-9',
       isSplit: true,
       budgetLines: [makeBudgetLine()],
     });
     const inv2 = makeInvoice({
       invoiceId: 'inv-2',
-      vendorName: 'Delta Corp',
-      invoiceNumber: 'D-1',
       isSplit: true,
       budgetLines: [makeBudgetLine()],
     });
-    const inv3 = makeInvoice({
-      invoiceId: 'inv-3',
-      vendorName: 'Epsilon Corp',
-      invoiceNumber: 'E-2',
-      isSplit: true,
-      budgetLines: [makeBudgetLine()],
-    });
-    const report = makeReport([inv1, inv2, inv3]);
-    const content = buildReportContent(
-      report,
-      new Set(['inv-1', 'inv-2', 'inv-3']),
-      'claim',
-      t,
-      formatters,
-    );
-    expect(content.footnotes).toEqual([
-      {
-        id: 'split',
-        marker: '†',
-        text: 'sourceReports.table.splitFootnote',
-      },
-    ]);
-    // Every split row carries the marker — no per-invoice numbering distinguishes them.
-    expect(content.rows.every((r) => r.allocatedMarkers === '†')).toBe(true);
-  });
-
-  it('AC1.3: produces no † marker and no split footnote anywhere when no included invoice is split', () => {
+    const report = makeReport([inv1, inv2]);
+    const content = buildReportContent(report, new Set(['inv-1', 'inv-2']), 'claim', t, formatters);
+    expect(content.footnotes).toEqual([]);
+    expect(content.rows.every((r) => r.isSplit)).toBe(true);
+  });
+
+  it('produces no footnotes when all invoices are unsplit', () => {
     const report = makeReport([makeInvoice({ isSplit: false })]);
     const content = buildReportContent(report, new Set(['inv-1']), 'claim', t, formatters);
     expect(content.footnotes).toEqual([]);
-    expect(content.rows[0]!.allocatedMarkers).toBe('');
+    expect(content.rows[0]!.isSplit).toBe(false);
   });
 
-  it('AC2.2: produces NO footnote entry for a constituted (tagged) deposit — the row gets isDeposit instead', () => {
+  it('produces no footnotes for constituted (tagged) deposit — the row gets isDeposit instead', () => {
     const invoice = makeInvoice({
       isSplit: true,
       budgetLines: [],
@@ -499,7 +481,7 @@ describe('buildReportContent — footnotes (AC1/AC2: shared, unnumbered, at most
     expect(content.rows[0]!.isDeposit).toBe(true);
   });
 
-  it('AC2.3: produces exactly one shared, unnumbered ‡ footnote when one or more invoices have a reduced (untagged) deposit', () => {
+  it('produces no footnotes when invoices have reduced (untagged) deposits — isDepositReduced is set instead', () => {
     const invoice = makeInvoice({
       isSplit: true,
       budgetLines: [],
@@ -507,35 +489,11 @@ describe('buildReportContent — footnotes (AC1/AC2: shared, unnumbered, at most
     });
     const report = makeReport([invoice], { id: 'src-1' });
     const content = buildReportContent(report, new Set(['inv-1']), 'claim', t, formatters);
-    expect(content.footnotes).toEqual([
-      {
-        id: 'deposit-reduced',
-        marker: '‡',
-        text: 'sourceReports.table.depositReducedFootnote',
-      },
-    ]);
-  });
-
-  it('AC2.3: multiple invoices with reduced deposits still produce exactly one shared ‡ entry', () => {
-    const inv1 = makeInvoice({
-      invoiceId: 'inv-1',
-      isSplit: true,
-      budgetLines: [],
-      deposits: [makeDeposit({ id: 'dep-1', budgetSourceId: null })],
-    });
-    const inv2 = makeInvoice({
-      invoiceId: 'inv-2',
-      isSplit: true,
-      budgetLines: [],
-      deposits: [makeDeposit({ id: 'dep-2', budgetSourceId: null })],
-    });
-    const report = makeReport([inv1, inv2], { id: 'src-1' });
-    const content = buildReportContent(report, new Set(['inv-1', 'inv-2']), 'claim', t, formatters);
-    const reducedFootnotes = content.footnotes.filter((f) => f.id === 'deposit-reduced');
-    expect(reducedFootnotes).toHaveLength(1);
+    expect(content.footnotes).toEqual([]);
+    expect(content.rows[0]!.isDepositReduced).toBe(true);
   });
 
-  it('AC2.4: both split and reduced-deposit invoices present → markers "†‡" on the combined row, footnotes ordered [split, deposit-reduced]', () => {
+  it('produces no footnotes when an invoice has both split lines and a reduced deposit', () => {
     const combined = makeInvoice({
       invoiceId: 'inv-1',
       isSplit: true,
@@ -544,9 +502,9 @@ describe('buildReportContent — footnotes (AC1/AC2: shared, unnumbered, at most
     });
     const report = makeReport([combined], { id: 'src-1' });
     const content = buildReportContent(report, new Set(['inv-1']), 'claim', t, formatters);
-    expect(content.rows[0]!.allocatedMarkers).toBe('†‡');
-    expect(content.footnotes.map((f) => f.id)).toEqual(['split', 'deposit-reduced']);
-    expect(content.footnotes).toHaveLength(2);
+    expect(content.footnotes).toEqual([]);
+    expect(content.rows[0]!.isSplit).toBe(true);
+    expect(content.rows[0]!.isDepositReduced).toBe(true);
   });
 
   it('produces no footnotes when no invoice is split or has a reduced deposit', () => {
diff --git a/client/src/lib/reportContent/buildReportContent.ts b/client/src/lib/reportContent/buildReportContent.ts
index a685718d1..811da696a 100644
--- a/client/src/lib/reportContent/buildReportContent.ts
+++ b/client/src/lib/reportContent/buildReportContent.ts
@@ -187,15 +187,8 @@ export function buildReportContent(
 
     const statusText = isOverview ? reportT(`sources.lines.invoiceStatus.${status}`) : null;
 
-    // Compute allocated markers († for split, ‡ for reduced; no markers for constituted deposits)
-    let allocatedMarkers = '';
-    if (splitInvoiceIds.has(invoice.invoiceId)) {
-      allocatedMarkers += '†';
-    }
-    if (depositReducedInvoiceIds.has(invoice.invoiceId)) {
-      allocatedMarkers += '‡';
-    }
-
+    const isSplit = splitInvoiceIds.has(invoice.invoiceId);
+    const isDepositReduced = depositReducedInvoiceIds.has(invoice.invoiceId);
     const isDeposit = depositConstitutedInvoiceIds.has(invoice.invoiceId);
     const refundNoteText = reportT('sourceReports.table.refundNote');
     const usageText = getUsageText(invoice);
@@ -211,7 +204,8 @@ export function buildReportContent(
       statusText,
       invoiceAmountText,
       allocatedAmountValueText,
-      allocatedMarkers,
+      isSplit,
+      isDepositReduced,
       isDeposit,
       isRefund: invoice.lineKind === 'refund-adjustment',
       refundNoteText,
@@ -235,25 +229,8 @@ export function buildReportContent(
     amountText: totalAmountText,
   });
 
-  // Build footnotes (at most 2 shared entries: split + deposit-reduced)
   const footnotes: ReportContentFootnote[] = [];
 
-  if (splitInvoiceIds.size > 0) {
-    footnotes.push({
-      id: 'split',
-      marker: '†',
-      text: reportT('sourceReports.table.splitFootnote'),
-    });
-  }
-
-  if (depositReducedInvoiceIds.size > 0) {
-    footnotes.push({
-      id: 'deposit-reduced',
-      marker: '‡',
-      text: reportT('sourceReports.table.depositReducedFootnote'),
-    });
-  }
-
   // Build cover letter (if enabled)
   let coverLetter: ReportContentCoverLetter | null = null;
   if (includeCoverLetter) {
@@ -298,6 +275,8 @@ export function buildReportContent(
       usage: reportT('sourceReports.table.usage'),
       attachmentsNote: reportT('sourceReports.editable.attachmentsNoteLabel'),
       deposit: reportT('sourceReports.table.attachmentType.deposit'),
+      splitNote: reportT('sourceReports.table.splitInlineLabel'),
+      depositReducedNote: reportT('sourceReports.table.depositReducedInlineLabel'),
       source: reportT('sourceReports.table.source'),
       sourceType: reportT('sourceReports.table.sourceType'),
       reference: reportT('sourceReports.table.reference'),
diff --git a/client/src/lib/reportContent/types.ts b/client/src/lib/reportContent/types.ts
index a7582a290..cc82f2388 100644
--- a/client/src/lib/reportContent/types.ts
+++ b/client/src/lib/reportContent/types.ts
@@ -13,8 +13,9 @@ export interface ReportContentRow {
   statusText: string | null; // null when useCase !== 'budget-overview'
   invoiceAmountText: string;
   allocatedAmountValueText: string; // formatted currency only — no markers/refund note
-  allocatedMarkers: string; // '', '†', '‡', '†‡' — shared/unnumbered per report
-  isDeposit: boolean; // constituted-deposit row → inline Deposit badge, no marker
+  isSplit: boolean; // split invoice with budget lines → inline "partial" label
+  isDepositReduced: boolean; // split invoice reduced by untagged deposits → inline label
+  isDeposit: boolean; // constituted-deposit row → inline Deposit badge
   isRefund: boolean;
   refundNoteText: string; // shown only when isRefund
   usageText: string; // EDITABLE — key `row.<invoiceId>.usageText`
@@ -55,6 +56,8 @@ export interface ReportContentLabels {
   usage: string;
   attachmentsNote: string;
   deposit: string; // translated in report language
+  splitNote: string; // short inline label for split rows
+  depositReducedNote: string; // short inline label for deposit-reduced rows
   source: string;
   sourceType: string;
   reference: string;
diff --git a/client/src/lib/reportPdf/coverLetterPdf.ts b/client/src/lib/reportPdf/coverLetterPdf.ts
index e61c2705b..5bdc9608c 100644
--- a/client/src/lib/reportPdf/coverLetterPdf.ts
+++ b/client/src/lib/reportPdf/coverLetterPdf.ts
@@ -56,12 +56,23 @@ export function buildCoverLetterContent(reportContent: ReportContent, t: TFuncti
     margin: [0, 0, 0, 16],
   });
 
-  // Body text — literal blank-line rendering, no paragraph-spacing model (see spec §B).
-  content.push({
-    text: coverLetter.body,
-    style: 'normal',
-    margin: [0, 0, 0, 32],
-  });
+  // Body text — split on double newlines so AI-generated paragraphs render with spacing
+  const paragraphs = coverLetter.body.split(/\n\n+/).filter(Boolean);
+  if (paragraphs.length <= 1) {
+    content.push({
+      text: coverLetter.body,
+      style: 'normal',
+      margin: [0, 0, 0, 32],
+    });
+  } else {
+    for (let i = 0; i < paragraphs.length; i++) {
+      content.push({
+        text: paragraphs[i]!,
+        style: 'normal',
+        margin: [0, 0, 0, i === paragraphs.length - 1 ? 32 : 8],
+      });
+    }
+  }
 
   // Signature block — closing + reserved blank space + name are ALWAYS emitted together (AC 2.4),
   // never gated behind `if (coverLetter.signature)`: pdfmake reserves the same line height for an
diff --git a/client/src/lib/reportPdf/overviewPdf.test.ts b/client/src/lib/reportPdf/overviewPdf.test.ts
index 213ddca5b..68e8d0744 100644
--- a/client/src/lib/reportPdf/overviewPdf.test.ts
+++ b/client/src/lib/reportPdf/overviewPdf.test.ts
@@ -73,7 +73,8 @@ function makeRow(overrides: Partial<ReportContentRow> = {}): ReportContentRow {
     statusText: null,
     invoiceAmountText: '€1000.00',
     allocatedAmountValueText: '€1000.00',
-    allocatedMarkers: '',
+    isSplit: false,
+    isDepositReduced: false,
     isDeposit: false,
     isRefund: false,
     refundNoteText: 'sourceReports.table.refundNote',
@@ -100,6 +101,8 @@ function makeLabels(): ReportContent['labels'] {
     usage: 'sourceReports.table.usage',
     attachmentsNote: 'sourceReports.editable.attachmentsNoteLabel',
     deposit: 'sourceReports.table.attachmentType.deposit',
+    splitNote: 'sourceReports.table.splitInlineLabel',
+    depositReducedNote: 'sourceReports.table.depositReducedInlineLabel',
     source: 'sourceReports.table.source',
     sourceType: 'sourceReports.table.sourceType',
     reference: 'sourceReports.table.reference',
@@ -767,8 +770,8 @@ describe('buildOverviewContent — row rendering (consumes already-derived Repor
     });
   });
 
-  describe('Usage cell: always a plain { text } cell — areaText/attachmentsNote render as SEPARATE continuation rows (#1929 round 4 architect review HIGH: the round-3 stack: [usageChunk, areaText, attachmentsNote] construction left their COMBINED height in one cell unbounded and silently dropped rows needing 3+/9+ pages; each field now gets its own independently-chunked row(s), never sharing a cell with usageText or with each other)', () => {
-    it('renders a plain { text } cell with no extra rows when both areaText and attachmentsNote are null', () => {
+  describe('Usage cell: plain text vs inline grey meta text', () => {
+    it('renders a plain { text } cell (not a stack) when both areaText and attachmentsNote are null', () => {
       const row = makeRow({ usageText: 'Kitchen work', areaText: null, attachmentsNote: null });
       const content = makeContent({ rows: [row] });
       const result = buildOverviewContent(content, new Map(), t);
@@ -780,27 +783,23 @@ describe('buildOverviewContent — row rendering (consumes already-derived Repor
       expect(table.body).toHaveLength(3);
     });
 
-    it('renders the usage row PLUS one continuation row (style "small") for attachmentsNote — never stacked into the usage cell', () => {
+    it('renders a text array with a grey newline run when attachmentsNote is present', () => {
       const row = makeRow({ usageText: 'Kitchen work', attachmentsNote: '1 attachment: Invoice' });
       const content = makeContent({ rows: [row] });
       const result = buildOverviewContent(content, new Map(), t);
       const table = getTable(result);
-
-      const usageCell = (table.body[1] as unknown[])[5] as { text?: unknown; stack?: unknown };
-      expect(usageCell.stack).toBeUndefined();
-      expect(usageRunsText(usageCell.text)).toBe('Kitchen work');
-
-      // header (1) + usage row (1) + attachmentsNote continuation row (1) + summary row (1) = 4.
-      expect(table.body).toHaveLength(4);
-      const noteRow = table.body[2] as { text?: unknown; style?: string }[];
-      // Leading/amount cells on the continuation row are all blank.
-      expect(rowTexts(noteRow).slice(0, 5)).toEqual(['', '', '', '', '']);
-      const noteCell = noteRow[5] as { text: unknown; style?: string };
-      expect(usageRunsText(noteCell.text)).toBe('1 attachment: Invoice');
-      expect(noteCell.style).toBe('small');
+      const cell = (table.body[1] as unknown[])[5] as {
+        text: { text: string; color?: string }[];
+        stack?: unknown;
+      };
+      expect(cell.stack).toBeUndefined();
+      expect(Array.isArray(cell.text)).toBe(true);
+      expect(cell.text[0]!.text).toBe('Kitchen work');
+      expect(cell.text[1]!.text).toContain('1 attachment: Invoice');
+      expect(cell.text[1]!.color).toBe('#6b7280');
     });
 
-    it('AC5.2: renders the usage row, then an areaText continuation row, then an attachmentsNote continuation row — in that order, each its own row', () => {
+    it('AC5.2: renders areaText and attachmentsNote joined in the grey meta run when both are present', () => {
       const row = makeRow({
         usageText: 'Kitchen work',
         areaText: 'Ground Floor',
@@ -809,20 +808,16 @@ describe('buildOverviewContent — row rendering (consumes already-derived Repor
       const content = makeContent({ rows: [row] });
       const result = buildOverviewContent(content, new Map(), t);
       const table = getTable(result);
-
-      // header (1) + usage row + areaText row + attachmentsNote row + summary row = 5.
-      expect(table.body).toHaveLength(5);
-      const usageCell = (table.body[1] as unknown[])[5] as { text: unknown };
-      const areaCell = (table.body[2] as unknown[])[5] as { text: unknown; style?: string };
-      const noteCell = (table.body[3] as unknown[])[5] as { text: unknown; style?: string };
-      expect(usageRunsText(usageCell.text)).toBe('Kitchen work');
-      expect(usageRunsText(areaCell.text)).toBe('Ground Floor');
-      expect(areaCell.style).toBe('small');
-      expect(usageRunsText(noteCell.text)).toBe('1 attachment: Invoice');
-      expect(noteCell.style).toBe('small');
+      const cell = (table.body[1] as unknown[])[5] as {
+        text: { text: string; color?: string }[];
+      };
+      expect(cell.text[0]!.text).toBe('Kitchen work');
+      expect(cell.text[1]!.text).toContain('Ground Floor');
+      expect(cell.text[1]!.text).toContain('1 attachment: Invoice');
+      expect(cell.text[1]!.color).toBe('#6b7280');
     });
 
-    it('renders the usage row plus only an areaText continuation row when areaText is present but attachmentsNote is null', () => {
+    it('renders areaText alone in the grey meta run when attachmentsNote is null', () => {
       const row = makeRow({
         usageText: 'Kitchen work',
         areaText: 'Ground Floor',
@@ -831,10 +826,14 @@ describe('buildOverviewContent — row rendering (consumes already-derived Repor
       const content = makeContent({ rows: [row] });
       const result = buildOverviewContent(content, new Map(), t);
       const table = getTable(result);
-      // header (1) + usage row + areaText row + summary row = 4.
-      expect(table.body).toHaveLength(4);
-      const areaCell = (table.body[2] as unknown[])[5] as { text: unknown };
-      expect(usageRunsText(areaCell.text)).toBe('Ground Floor');
+      const cell = (table.body[1] as unknown[])[5] as {
+        text: { text: string; color?: string }[];
+        stack?: unknown;
+      };
+      expect(cell.stack).toBeUndefined();
+      expect(cell.text[0]!.text).toBe('Kitchen work');
+      expect(cell.text[1]!.text).toContain('Ground Floor');
+      expect(cell.text[1]!.color).toBe('#6b7280');
     });
 
     it('[#1929 round 2] the plain-cell Usage text is a run array of the individual whitespace-preserving tokens (buildUsageTextRuns wiring, not a plain string)', () => {
@@ -848,34 +847,40 @@ describe('buildOverviewContent — row rendering (consumes already-derived Repor
     });
   });
 
-  describe('allocated cell composition (skip markers + allocatedMarkers + refund note)', () => {
+  describe('allocated cell composition (skip markers + inline labels + refund note)', () => {
     it('renders allocatedAmountValueText plain when there are no markers and not a refund', () => {
-      const row = makeRow({ allocatedAmountValueText: '€400.00', allocatedMarkers: '' });
+      const row = makeRow({ allocatedAmountValueText: '€400.00' });
       const content = makeContent({ rows: [row] });
       const result = buildOverviewContent(content, new Map(), t);
       const table = getTable(result);
       expect(rowTexts(table.body[1])[4]).toBe('€400.00');
     });
 
-    it('appends the pre-computed, unnumbered/shared split+deposit markers verbatim (already formatted by buildReportContent)', () => {
-      const row = makeRow({ allocatedAmountValueText: '€400.00', allocatedMarkers: '†‡' });
+    it('appends inline isSplit label when isSplit=true', () => {
+      const row = makeRow({ allocatedAmountValueText: '€400.00', isSplit: true });
       const content = makeContent({ rows: [row] });
       const result = buildOverviewContent(content, new Map(), t);
       const table = getTable(result);
-      expect(rowTexts(table.body[1])[4]).toBe('€400.00†‡');
+      expect(rowTexts(table.body[1])[4]).toContain('€400.00');
+      expect(rowTexts(table.body[1])[4]).toContain('sourceReports.table.splitInlineLabel');
     });
 
-    it('prepends skip-footnote markers (*N) BEFORE the allocatedMarkers, numbered from skippedDocuments', () => {
-      const row = makeRow({
-        invoiceId: 'inv-1',
-        allocatedAmountValueText: '€400.00',
-        allocatedMarkers: '†',
-      });
+    it('appends inline isDepositReduced label when isDepositReduced=true', () => {
+      const row = makeRow({ allocatedAmountValueText: '€400.00', isDepositReduced: true });
+      const content = makeContent({ rows: [row] });
+      const result = buildOverviewContent(content, new Map(), t);
+      const table = getTable(result);
+      expect(rowTexts(table.body[1])[4]).toContain('€400.00');
+      expect(rowTexts(table.body[1])[4]).toContain('sourceReports.table.depositReducedInlineLabel');
+    });
+
+    it('prepends skip-footnote markers (*N) before the allocated value, numbered from skippedDocuments', () => {
+      const row = makeRow({ invoiceId: 'inv-1', allocatedAmountValueText: '€400.00' });
       const content = makeContent({ rows: [row] });
       const skipped = new Map<string, string[]>([['inv-1', ['footnoteFetchFailed']]]);
       const result = buildOverviewContent(content, skipped, t);
       const table = getTable(result);
-      expect(rowTexts(table.body[1])[4]).toBe('€400.00*1†');
+      expect(rowTexts(table.body[1])[4]).toBe('€400.00*1');
     });
 
     it('numbers multiple skip reasons on the same invoice sequentially', () => {
diff --git a/client/src/lib/reportPdf/overviewPdf.ts b/client/src/lib/reportPdf/overviewPdf.ts
index b51a8cfc1..02aa870d3 100644
--- a/client/src/lib/reportPdf/overviewPdf.ts
+++ b/client/src/lib/reportPdf/overviewPdf.ts
@@ -581,45 +581,18 @@ export function buildOverviewContent(
   const usageSafeTokenChars = reportContent.isOverview
     ? USAGE_SAFE_TOKEN_CHARS_7COL
     : USAGE_SAFE_TOKEN_CHARS_6COL;
-  const smallSafeTokenChars = reportContent.isOverview
-    ? SMALL_SAFE_TOKEN_CHARS_7COL
-    : SMALL_SAFE_TOKEN_CHARS_6COL;
-
-  /**
-   * Pushes one continuation row (empty leading/amount cells) per chunk of `text`, at `style`.
-   * Used for usageText overflow, and — separately, never combined with usageText or with each
-   * other — for areaText/attachmentsNote (#1929 round-4 architect review HIGH: each field gets
-   * its own row(s) so no row's Usage cell ever holds more than one bounded chunk of one field;
-   * see MAX_SAFE_USAGE_CHUNK_CHARS's comment for what "bounded" means and how it was measured).
-   */
-  function pushChunkedRows(
-    text: string,
-    maxChunkChars: number,
-    safeTokenCharsForStyle: number,
-    style: 'tableCell' | 'small',
-  ): void {
-    for (const chunk of splitIntoPageSafeChunks(text, maxChunkChars)) {
-      const cell: Content = { text: buildUsageTextRuns(chunk, safeTokenCharsForStyle), style };
-      rows.push([
-        ...buildEmptyLeadingCells(reportContent.isOverview),
-        ...buildEmptyAmountCells(),
-        cell,
-      ]);
-    }
-  }
 
   for (const contentRow of reportContent.rows) {
-    // Allocated amount with footnote markers (skip + allocated)
+    // Allocated amount with skip markers and inline labels
     const skipMarkers = skipFootnotesByInvoiceId.get(contentRow.invoiceId) ?? [];
-    let markerText = '';
+    let skipMarkerText = '';
     for (const noteNum of skipMarkers) {
-      markerText += `*${noteNum}`;
+      skipMarkerText += `*${noteNum}`;
     }
-    markerText += contentRow.allocatedMarkers;
 
-    // Build allocated runs: value+markers, then optional deposit badge, then optional refund note
+    // Build allocated runs: value+skip markers, then optional inline labels, then optional refund note
     const allocatedRuns: Content[] = [
-      { text: `${contentRow.allocatedAmountValueText}${markerText}` },
+      { text: `${contentRow.allocatedAmountValueText}${skipMarkerText}` },
     ];
     if (contentRow.isDeposit) {
       allocatedRuns.push({
@@ -628,6 +601,20 @@ export function buildOverviewContent(
         fontSize: DEPOSIT_NOTE_FONT_SIZE,
       });
     }
+    if (contentRow.isSplit) {
+      allocatedRuns.push({
+        text: ` (${reportContent.labels.splitNote})`,
+        color: DEPOSIT_NOTE_TEXT_COLOR,
+        fontSize: DEPOSIT_NOTE_FONT_SIZE,
+      });
+    }
+    if (contentRow.isDepositReduced) {
+      allocatedRuns.push({
+        text: ` (${reportContent.labels.depositReducedNote})`,
+        color: DEPOSIT_NOTE_TEXT_COLOR,
+        fontSize: DEPOSIT_NOTE_FONT_SIZE,
+      });
+    }
     if (contentRow.isRefund) {
       allocatedRuns.push({ text: ` ${contentRow.refundNoteText}` });
     }
@@ -637,13 +624,18 @@ export function buildOverviewContent(
     // table rows instead of one unbreakable (and potentially content-dropping) row. The FIRST
     // chunk shares this invoice's leading/amount-cell row; any further chunks (rare-by-
     // construction: AC12 requires 600 chars with zero degradation, well under
-    // MAX_SAFE_USAGE_CHUNK_CHARS) get their own continuation row via pushChunkedRows — no
-    // "continued" marker, per the product-owner's explicit ruling.
+    // MAX_SAFE_USAGE_CHUNK_CHARS) get their own continuation row — no "continued" marker,
+    // per the product-owner's explicit ruling.
     const usageChunks = splitIntoPageSafeChunks(contentRow.usageText, MAX_SAFE_USAGE_CHUNK_CHARS);
-    const firstUsageCell: Content = {
-      text: buildUsageTextRuns(usageChunks[0]!, usageSafeTokenChars),
-      style: 'tableCell',
-    };
+    // Area and attachmentsNote appear inline in grey after the first usage chunk
+    const metaPieces: string[] = [];
+    if (contentRow.areaText) metaPieces.push(contentRow.areaText);
+    if (contentRow.attachmentsNote) metaPieces.push(contentRow.attachmentsNote);
+    const firstUsageRuns = buildUsageTextRuns(usageChunks[0]!, usageSafeTokenChars);
+    if (metaPieces.length > 0) {
+      firstUsageRuns.push({ text: '\n' + metaPieces.join(' · '), color: DEPOSIT_NOTE_TEXT_COLOR });
+    }
+    const firstUsageCell: Content = { text: firstUsageRuns, style: 'tableCell' };
     rows.push([
       ...buildLeadingCells(contentRow, reportContent.isOverview, contentRow.statusText ?? ''),
       ...buildAmountCells(contentRow, allocatedRuns),
@@ -660,28 +652,6 @@ export function buildOverviewContent(
         cell,
       ]);
     }
-
-    // areaText and attachmentsNote each get their OWN continuation row(s) — never stacked into
-    // the usage row's cell (#1929 round-4 architect review HIGH: attachmentsNote has no
-    // maxLength anywhere, and areaText is aggregate-unbounded across N leaf areas, so their
-    // combined height with usageText in one cell was unbounded and silently dropped rows that
-    // needed more than one page).
-    if (contentRow.areaText) {
-      pushChunkedRows(
-        contentRow.areaText,
-        MAX_SAFE_SMALL_CHUNK_CHARS,
-        smallSafeTokenChars,
-        'small',
-      );
-    }
-    if (contentRow.attachmentsNote) {
-      pushChunkedRows(
-        contentRow.attachmentsNote,
-        MAX_SAFE_SMALL_CHUNK_CHARS,
-        smallSafeTokenChars,
-        'small',
-      );
-    }
   }
 
   // Add summary rows from reportContent.summaryRows
@@ -727,7 +697,7 @@ export function buildOverviewContent(
     margin: [0, 0, 0, 20],
   });
 
-  // Add footnotes (skip block + split/deposit from reportContent.footnotes)
+  // Add footnotes (skip block only; split/deposit annotations are now rendered inline)
   const footnotes: Content[] = [];
 
   // Skip block (generation-time data)
diff --git a/client/src/lib/reportPdf/realRender.test.ts b/client/src/lib/reportPdf/realRender.test.ts
index e6bb190dd..9b7968262 100644
--- a/client/src/lib/reportPdf/realRender.test.ts
+++ b/client/src/lib/reportPdf/realRender.test.ts
@@ -914,14 +914,14 @@ describe('report PDF pipeline — real, unmocked end-to-end render', () => {
           household: null,
         });
 
-        // Sanity: the constituted-deposit row carries isDeposit=true and no ‡ marker; the
-        // reduced-deposit row carries the ‡ marker and isDeposit=false.
+        // Sanity: the constituted-deposit row carries isDeposit=true; the
+        // reduced-deposit row carries isDepositReduced=true and isDeposit=false.
         const constitutedRow = content.rows.find((r) => r.invoiceId === 'inv-deposit-constituted')!;
         expect(constitutedRow.isDeposit).toBe(true);
-        expect(constitutedRow.allocatedMarkers).not.toContain('‡');
+        expect(constitutedRow.isDepositReduced).toBe(false);
         const reducedRow = content.rows.find((r) => r.invoiceId === 'inv-deposit-reduced')!;
         expect(reducedRow.isDeposit).toBe(false);
-        expect(reducedRow.allocatedMarkers).toContain('‡');
+        expect(reducedRow.isDepositReduced).toBe(true);
 
         const pdfContent = buildOverviewContent(content, new Map(), t);
         const tableItem = pdfContent.find(

From 9eaa92955eccd9fb6cd66e77475e2e5d6d1f0ef5 Mon Sep 17 00:00:00 2001
From: Frank Steiler <frank@steiler.dev>
Date: Mon, 3 Aug 2026 18:10:21 +0200
Subject: [PATCH 2/6] fix(reports): bound the inline usage meta so the PDF
 stops dropping content
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

The inline-meta change appended areaText/attachmentsNote as one unchunked
run on the usage cell. With table.dontBreakRows, pdfmake does not
paginate an over-tall row — it silently discards the overflow. This
reintroduced the #1929 round-4 content-loss regression: page count
saturated at 2 and went non-monotonic (3 -> 2 as content grew) while
rendered line count kept rising linearly, so ~7 pages of measured
content were thrown away. Reachable from ordinary data: attachmentsNote
has no maxLength and areaText is unbounded across leaf areas.

packUsageCellRows() now bounds the whole rendered cell stream rather than
usageText alone, splitting via the existing splitIntoPageSafeChunks().
Page count is monotonic again and identical whichever channel carries the
text. Where the cell fits one page-safe row — the dominant case — output
is byte-identical.

The grey suffix now lands on the last chunk row rather than the first:
pinning it to row 0 rendered meta mid-prose with more usage below it, and
forced the usage chunk boundary to shrink.

Also updates the tests that encoded the unbounded model. Two of them
passed while the PDF lost content, because they inspected the pdfmake
content tree rather than rendered output; the guard is now framed as
rendered-output-only and asserts that the same text costs the same number
of pages whichever channel carries it. Adds coverage for packUsageCellRows
(none existed), the inline split/deposit labels, and the column toggles.
Removes the small-chunk constants, now dead.

E2E: reportWizard scenarios updated for the renamed usage-meta class, the
removed attachments field, and the footnote-to-inline-label change.

German strings verified against the glossary; abzgl. Abschlag kept over
the approved Abschlagszahlung because the compliant term measures 96pt
against a 75pt column.

Co-Authored-By: Claude frontend-developer <noreply@anthropic.com>
Co-Authored-By: Claude qa-integration-tester <noreply@anthropic.com>
Co-Authored-By: Claude e2e-test-engineer <noreply@anthropic.com>
Co-Authored-By: Claude translator <noreply@anthropic.com>
---
 .../agent-memory/e2e-test-engineer/MEMORY.md  |   1 +
 .../issue-1959-inline-meta-and-labels.md      |  55 ++
 .../agent-memory/frontend-developer/MEMORY.md |   1 +
 .../frontend-developer/qa-tripwire-tests.md   |  30 +
 .../qa-integration-tester/MEMORY.md           |   1 +
 .../pr-1959-inline-meta-content-loss.md       | 114 ++++
 .../reports/ReportContentEditor.test.tsx      | 482 +++++++++++++--
 client/src/lib/reportContent/types.ts         |   6 +-
 .../src/lib/reportPdf/coverLetterPdf.test.ts  |   2 +
 client/src/lib/reportPdf/merge.test.ts        |   2 +
 client/src/lib/reportPdf/overviewPdf.test.ts  | 442 +++++++++++---
 client/src/lib/reportPdf/overviewPdf.ts       | 274 ++++++---
 client/src/lib/reportPdf/pageGeometry.test.ts |   9 +-
 client/src/lib/reportPdf/pageGeometry.ts      |  12 +-
 client/src/lib/reportPdf/realRender.test.ts   | 572 ++++++++++++++----
 e2e/pages/ReportWizardPage.ts                 | 134 ++--
 .../budget/reportWizardAiGeneration.spec.ts   |  32 +-
 .../reportWizardEditableContent.spec.ts       | 235 +++++--
 18 files changed, 1909 insertions(+), 495 deletions(-)
 create mode 100644 .claude/agent-memory/e2e-test-engineer/issue-1959-inline-meta-and-labels.md
 create mode 100644 .claude/agent-memory/frontend-developer/qa-tripwire-tests.md
 create mode 100644 .claude/agent-memory/qa-integration-tester/pr-1959-inline-meta-content-loss.md

diff --git a/.claude/agent-memory/e2e-test-engineer/MEMORY.md b/.claude/agent-memory/e2e-test-engineer/MEMORY.md
index dcd665f33..c2037cf37 100644
--- a/.claude/agent-memory/e2e-test-engineer/MEMORY.md
+++ b/.claude/agent-memory/e2e-test-engineer/MEMORY.md
@@ -29,6 +29,7 @@
 - [story-1901-ai-report-generation.md](story-1901-ai-report-generation.md) — AI-generated usage descriptions/cover letter: new `reportWizardAiGeneration.spec.ts` + POM AI locators; llmEnabled-mock pattern; overwrite-modal-guard-checks-overrides-not-aiContent gotcha; cover-letter-fields-need-contactAddress/reference seed trap; reused auto-itemize LLM error strings.
 - [story-1900-editable-report-preview.md](story-1900-editable-report-preview.md) — step 5 reworked from always-present auto-regenerating iframe → live editable surface + on-demand PDF Modal; POM rename `waitForPreviewReady/Regenerated` → `openPdfPreviewModal`/`closePdfPreviewModal` (must close before another modal-opening action); 4 filed bugs (#1904-#1907, all now FIXED+CLOSED, see file's re-verification section); deliberate deviation editing a file explicitly marked "do not touch" because leaving it broke `tsc` workspace-wide — see file for the general lesson; `footnoteFetchFailed` skip note naturally reachable with no Paperless container, no mock needed; re-verification added `mobileCard()`/`mobileUsageField()` POM locators + Scenario 15 (#1907 regression guard) + found/filed NEW bug #1908 (mobile-card fallback visible on desktop, no default `display:none`) with its own expected-to-fail Scenario 1b.
 
+- [issue-1959-inline-meta-and-labels.md](issue-1959-inline-meta-and-labels.md) — #1959 reversed #1923's †/‡ footnotes → inline `(partial)`/`(less deposit)` and merged area+attachments into one `.usageMetaText` line; POM renames, rewritten scenarios 2/17/18/20 + AI 8, attachment-tier facts, untested column toggles.
 - [claim-deposit-scope-1922.md](claim-deposit-scope-1922.md) — PR #1922 invoice/deposit claim-scope split: `handleMarkClaimed`'s two-array submit, server-truth success-banner counts, the three "deposit surfaces the invoice" shapes, `claimNothingClaimable` guard.
 
 ## Open follow-ups to flag to orchestrator
diff --git a/.claude/agent-memory/e2e-test-engineer/issue-1959-inline-meta-and-labels.md b/.claude/agent-memory/e2e-test-engineer/issue-1959-inline-meta-and-labels.md
new file mode 100644
index 000000000..5375285b2
--- /dev/null
+++ b/.claude/agent-memory/e2e-test-engineer/issue-1959-inline-meta-and-labels.md
@@ -0,0 +1,55 @@
+---
+name: issue-1959-inline-meta-and-labels
+description: PR #1959 reversed two earlier report-table designs (†/‡ shared footnotes from #1923, distinct area sub-line) into inline labels + one combined meta line; which E2E locators/scenarios had to be rewritten and how each new assertion was made non-vacuous.
+metadata:
+  type: project
+---
+
+PR #1959 ("improve report PDF UX") deliberately **superseded** two designs earlier rounds had
+asked for, in `ReportContentEditor.tsx` / `buildReportContent.ts`:
+
+1. `†`/`‡` markers + the shared footnote list (Story #1923 AC1) → grey inline `<span
+   class*="inlineNote">` in the **Allocated Amount cell**: `(partial)` / `(less deposit)`
+   (de `(Teilbetrag)` / `(abzgl. Abschlag)`). `ReportContentRow.allocatedMarkers` → `isSplit` /
+   `isDepositReduced` booleans. `buildReportContent` now pushes **zero** footnotes, so
+   `.footnotes` has no producer at all — `footnotesBlock`/`footnoteItems` survive in the POM as
+   **negative-only** guards.
+2. `.usageAreaText` sub-line + the separate editable `Attachments Note` column → ONE read-only
+   `.usageMetaText` line inside the Usage cell: `[areaText, attachmentsNote].join(' · ')`
+   (U+00B7 middle dot, spaces on both sides). The `attachmentsNote` `EditableField` is gone
+   entirely, so a content-table row / mobile card now has **exactly one textbox** (Usage) — the
+   crispest available guard against that column coming back.
+
+**Why:** the user owns #1959 and asked for it in the promotion; source-of-truth hierarchy makes
+the PR body the spec, so the tests were rewritten, not the code.
+
+**How to apply:** POM renames are `usageAreaText`→`usageMetaText`,
+`mobileUsageAreaText`→`mobileUsageMetaText`, plus new `inlineNote()`/`mobileInlineNote()`;
+`attachmentsNoteField()` deleted. Rewritten scenarios: editableContent 2, 17, 18, 20 and
+aiGeneration 8. Every "old design is gone" negative is paired with a positive so it cannot pass
+against a mis-seeded page (e.g. Scenario 18 asserts `(partial)` present *and* `†`/`‡` absent
+*and* the long-form footnote sentence absent from `main`; Scenario 20 asserts the attachments
+note text IS rendered *and* the row has one textbox).
+
+Facts worth reusing:
+
+- `toHaveText`/`toContainText` normalize whitespace, so the desktop `inlineNote` span's leading
+  space (` (partial)`) is absorbed — `toHaveText('(partial)')` is correct.
+- Attachment note text for a `claim` report with `attachmentType: 'invoice'` is
+  `1 attachment: Invoice`. The tier gate is
+  `server/src/services/shared/attachmentTierUtils.ts`: floors are quotation(1) for
+  budget-overview, deposit(2) for claim, invoice(3) for proof-of-funds; `null` counts as tier 3.
+  So a 'invoice'-tagged link shows up in claim reports — don't guess, that file is the only
+  definition.
+- `document_links` is unique on `(entity_type, entity_id, paperless_document_id)`, so a
+  hardcoded `paperlessDocumentId` in a spec cannot collide across parallel workers/projects.
+- `AppShell.tsx` renders a real `<main>` element, so `page.locator('main')` is a safe
+  page-scope text container.
+- **Column-visibility checkboxes (`role="group"`, "Show/hide columns") shipped with #1959 with
+  NO E2E coverage** — deliberately not added on the critical path, because a new test case
+  reshuffles shard membership. Pick this up when the promotion isn't blocking.
+- Sub-agents share the worktree here: a `prettier --check` on a `client/` file can transiently
+  fail because another agent is mid-write. Re-check before reporting it as broken.
+
+See [[known-flakes-and-regressions]], [[story-1900-editable-report-preview]],
+[[story-1879-report-wizard]].
diff --git a/.claude/agent-memory/frontend-developer/MEMORY.md b/.claude/agent-memory/frontend-developer/MEMORY.md
index f8666cdbf..aed9b8519 100644
--- a/.claude/agent-memory/frontend-developer/MEMORY.md
+++ b/.claude/agent-memory/frontend-developer/MEMORY.md
@@ -98,6 +98,7 @@ removing it will fail typecheck. Keep the export and note it as QA-deferred.
 - `budget-invoices.md` — open for budget/invoice/subsidy pages: BudgetSubNav, budget line APIs, invoice CRUD patterns, pipeline dashboard cards
 - `gantt-calendar.md` — open for Timeline/Gantt/Calendar/Milestone work: SVG architecture, drag hook (React 19 refs rule), arrow hover, lane allocation, constraints section, related E2E POM layout
 - `api-routing.md` — open when calling APIs or navigating: UserResponse/PaginatedResponse shapes, auth routes, dependency direction-swap semantics, navigation origin state
+- `qa-tripwire-tests.md` — open when a task says a fix will flip an `it.failing` guard: why the flip signal can be swallowed by a shared helper, and what to do instead
 - `testing-mocks.md` — open when your change will break existing tests: ESM mock patterns, mock-block update checklist, Toast/context mocks, DefinePlugin globals in Jest
 - `git-tooling.md` — open on webpack config changes or git/CI trouble: worktree remotes, object corruption/permission recovery, npm audit policy, Prettier/CI format gap
 - `photo-metadata-sidepanel.md` — photo metadata side panel notes (open for photo UI work)
diff --git a/.claude/agent-memory/frontend-developer/qa-tripwire-tests.md b/.claude/agent-memory/frontend-developer/qa-tripwire-tests.md
new file mode 100644
index 000000000..14fff6f1d
--- /dev/null
+++ b/.claude/agent-memory/frontend-developer/qa-tripwire-tests.md
@@ -0,0 +1,30 @@
+---
+name: qa-tripwire-tests
+description: QA encodes known production defects as `it.failing` tripwires; a shared helper that hard-codes the broken model can swallow the flip signal, so verify the fix independently
+metadata:
+  type: project
+---
+
+QA (`qa-integration-tester`) encodes a known-but-unfixed production defect as `it.failing` with a
+passing control test alongside. The suite stays green while the defect is open; when production is
+fixed the test starts passing and Jest errors with "Failing test passed even though it was supposed
+to fail" — that error is the intended confirmation signal, and the fixing agent is expected to flip
+`it.failing` → `it` (the one sanctioned developer edit to a test file).
+
+**Why:** it keeps the defect encoded in executable code rather than only in a report, so a fix can't
+land without someone noticing the guard.
+
+**How to apply:** the flip signal is NOT reliable on its own. If the tripwire calls a shared render
+helper that asserts the _broken_ model (e.g. `expect(body).toHaveLength(1 + usageChunkCount + 1)`,
+which bakes in "meta never adds rows"), a correct fix makes the helper throw first — the tripwire
+then still "passes" as a failing test and no signal appears. Seen on #1959 (`realRender.test.ts`
+cell-scope block): the fix was verified, the tripwire never flipped.
+
+So: (1) always verify the fix independently with your own scratch render/measurement harness before
+trusting or distrusting the tripwire; (2) if the flip leaves the test failing inside a shared helper
+rather than at its own assertions, revert to `it.failing`, leave QA's file byte-identical, and report
+the exact helper line QA must update — do not "fix" the helper to make the flip work;
+(3) expect sibling tests in the same block to fail too, for the same reason (they assert the broken
+model at content-tree level, which is why they passed while the rendered PDF lost content).
+
+Related: [[../MEMORY.md]] "Refinement Workflow — QA Test Coordination".
diff --git a/.claude/agent-memory/qa-integration-tester/MEMORY.md b/.claude/agent-memory/qa-integration-tester/MEMORY.md
index a42ecf663..0a27ee35a 100644
--- a/.claude/agent-memory/qa-integration-tester/MEMORY.md
+++ b/.claude/agent-memory/qa-integration-tester/MEMORY.md
@@ -15,6 +15,7 @@
 
 ## Recent bug/story notes (2026-08)
 
+- [PR #1959 — inline meta content loss + `it.failing` tripwires](pr-1959-inline-meta-content-loss.md) (2026-08-03, RESOLVED) — prod defect found+fixed (unchunked meta in a `dontBreakRows` cell silently drops pages); **a tripwire is worthless if a shared helper bakes in the buggy assumption** — mine nearly stayed green through the fix; channel-equivalence is the threshold-free assertion; tree-level assertions cannot see this bug class; `grep` silently returns nothing on these test files (use `awk`).
 - [Bug #1955 — echo-race harness + mutation probes](bug-1955-echo-race-harness.md) (2026-08-03) — echo must fire on the write's *resolve* (not the call) or the queue fix masks the guard and the test passes pre-fix; `rerender()` stands in for the optimistic `setPreferences`; 4 perl mutation probes prove each test guards a distinct part of the fix; never run repo-wide `npm run format` (38 unrelated files drift).
 
 - [Story #1930 — attachment tier rules](story-1930-attachment-tier.md) (2026-08-02) — new `attachmentTierUtils.test.ts` (100% cov, table+explicit-null pattern); sourceReportService.test.ts 16a/b/c/e → AC1/AC2/AC3/AC5; cross-report-type test needs a fresh invoice per report type; round 2 added a genuinely discriminating proof-of-funds fixture (deposit-only invoice, tagged+claimed deposit) after PO flagged the original AC3/16 pof blocks weren't change-detecting — inline-swap isolation technique to prove it fails on old logic.
diff --git a/.claude/agent-memory/qa-integration-tester/pr-1959-inline-meta-content-loss.md b/.claude/agent-memory/qa-integration-tester/pr-1959-inline-meta-content-loss.md
new file mode 100644
index 000000000..4f76d8f57
--- /dev/null
+++ b/.claude/agent-memory/qa-integration-tester/pr-1959-inline-meta-content-loss.md
@@ -0,0 +1,114 @@
+---
+name: pr-1959-inline-meta-content-loss
+description: PR #1959 (report PDF UX) reintroduced and then fixed the #1929 round-4 PDF content-loss regression; records the packUsageCellRows bound, the measured page-count saturation table, the it.failing tripwire technique AND the trap that nearly disarmed it (a helper that bakes in the buggy assumption), and the run-index gotcha for grey meta runs
+metadata:
+  type: project
+---
+
+# PR #1959 — inline meta content loss: the defect, the fix, and the tripwire trap
+
+## Status: RESOLVED (2026-08-03, branch `fix/1959-pdf-ux-green`)
+
+Production is fixed; the guard tests are normal `it` again. Everything below is the reusable lesson.
+
+## The defect
+
+`overviewPdf.ts` `buildOverviewContent` appended `areaText`/`attachmentsNote` as ONE **unchunked**
+grey run on a usage-chunk row's cell. With `table.dontBreakRows: true` pdfmake does not paginate an
+over-tall row — it measures the text and then **silently drops** whatever doesn't fit. #1929 round 4
+had bounded exactly this by giving each field its own chunked continuation row; #1959's inline move
+removed the bound without replacing it.
+
+**Why it happened:** the PR body asked for inline rendering and the implementer took it literally
+without re-bounding the *combined* cell height. `attachmentsNote` has no maxLength anywhere (editor
+or server) and `areaText` is aggregate-unbounded across N leaf areas, so ordinary user data reached it.
+
+**Measured evidence** (real unmocked renders, production 6-column shape). Rendered *line* count kept
+growing linearly (~0.023 lines/char) while page count saturated — proof pdfmake measured everything
+then discarded it:
+
+| chars | usageText (chunked) | attachmentsNote — before fix | after fix |
+| --- | --- | --- | --- |
+| 2000 | 6 rows / 2 pages | 3 rows / 3 pages | 6 rows / 2 pages |
+| 4000 | 9 rows / 3 pages | 3 rows / **2** pages | 9 rows / 3 pages |
+| 8000 | 15 rows / 5 pages | 3 rows / **2** pages | 15 rows / 5 pages |
+| 16000 | 27 rows / 9 pages | 3 rows / **2** pages | 27 rows / 9 pages |
+
+Page count was even non-monotonic (3 → 2 as content grew). Reproduces the architect's original tell
+verbatim: "rows requiring 3 and 9 pages both rendered as 2".
+
+## The fix, and what it changed for tests
+
+`packUsageCellRows(segments, maxChars)` packs the cell's **whole** content stream (prose, then the
+grey meta segment) into per-row groups of ≤ `MAX_SAFE_USAGE_CHUNK_CHARS`. Lossless; byte-identical
+output when the whole cell fits one row (the dominant case).
+
+Two consequences tests must encode, not work around:
+
+1. **The suffix now lands on the LAST row, not the first.** Pinning it to row 0 rendered grey meta
+   mid-prose with more usage below it, and shrank the prose's chunk boundary (adding rows).
+2. **The suffix can span many rows** — one grey run per row, always last within its own cell. So
+   reconstructing it means concatenating the grey run of *every* row in the group.
+
+`MAX_SAFE_SMALL_CHUNK_CHARS` and `SMALL_SAFE_TOKEN_CHARS_7COL/_6COL` were 9pt ceilings for the
+continuation rows that no longer exist; the meta renders at 8pt `tableCell`, so the usage budget is
+the right one. All three (plus the orphaned `SMALL_WORST_CASE_CHAR_WIDTH_PT`) were deleted along with
+their imports and value-only tests. `TABLE_SMALL_FONT_SIZE` stays — `PDF_STYLES.small` still renders
+footnotes, cover-letter date/reference lines, and the running header/footer.
+
+## Technique: `it.failing` as a tripwire for a known-open production defect
+
+Jest 30 here supports `it.failing` / `test.failing`. It is the right primitive when QA must leave a
+suite green but must not bless broken behaviour: the suite reports **passed** while production is
+broken, and the moment production is fixed Jest errors with `Failing test passed even though it was
+supposed to fail`, forcing a flip back to `it`. Verify the guard is live by temporarily relaxing its
+threshold and confirming that error appears.
+
+### The trap that nearly disarmed it — check this every time
+
+**A tripwire is worthless if a shared helper bakes in the very assumption the bug lives in.** Here
+`renderCellScopeRow` derived its expected row count from `usageText` alone — i.e. it asserted "the
+meta never adds rows", which is precisely what the bug did. Any correct fix *must* add rows, so the
+helper threw on its own row-count assertion before the test reached its page-count assertion: the
+test failed for the wrong reason and `it.failing` stayed green through the fix.
+
+Rules that follow:
+- Derive expected shape from the **same function production uses** (here `packUsageCellRows`), never
+  from a re-derivation of one input channel.
+- Assert losslessness and budget bounds against the **INPUT and the declared constant**, never
+  against the packer's own output — otherwise a packing regression satisfies them by moving in step.
+- After flipping a tripwire, confirm it passes **on its own assertion**, not merely that the suite is
+  green.
+
+## Strongest formulation found: channel equivalence
+
+The assertion that needed no calibrated threshold and is immune to future layout tuning:
+**the same text costs the same number of pages whichever channel carries it** (`usageText` vs
+`attachmentsNote`). Pre-fix this failed spectacularly (9 pages vs 2 at 16,000 chars). Pair it with a
+per-channel losslessness check first, so "same page count" can't be satisfied by both channels
+dropping equally.
+
+Related: tree-level assertions **cannot see this bug class at all** — the dropped text is present in
+the pdfmake content tree while the reader receives a truncated PDF. Where the claim is about what the
+reader receives, assert against rendered output (page count).
+
+## Gotcha: the grey meta run is never at a fixed run index
+
+`buildUsageTextRuns()` tokenizes prose into **one run per whitespace-delimited token**, so a cell's
+`.text` for `'Kitchen work'` is `[{text:'Kitchen'},{text:' '},{text:'work'}]` and the meta run is at
+`text[3]`, not `text[1]`. Locate it by its colour (`#6b7280` = `DEPOSIT_NOTE_TEXT_COLOR`), assert it
+is **last within its cell**, and reconstruct prose from the runs before it — see `splitUsageCell()`
+in both `overviewPdf.test.ts` and `realRender.test.ts`. Note the production code strips the leading
+`'\n'` when the suffix *starts* a cell (its own continuation row), so keep both a raw and a stripped
+accessor for faithful cross-row concatenation.
+
+## Environment gotcha: `grep` silently returns nothing on these test files
+
+After `npx prettier --write`, `grep` (even `grep -c ""`) on `overviewPdf.test.ts` and
+`ReportContentEditor.test.tsx` returns **no output and no error** — binary-content detection on some
+byte sequence. This produced a false "constant is unused" conclusion mid-task. Use
+`awk '/pattern/{print NR": "$0}' file` instead whenever a grep result on a report-PDF test file looks
+suspiciously empty. `npx eslint` is the reliable authority on unused imports.
+
+Related: [[test-infra-reference]], [[story-1929-round2-real-render-technique]],
+[[story-1923-report-table-cleanup]], [[story-1898-report-table-refinements]]
diff --git a/client/src/components/reports/ReportContentEditor.test.tsx b/client/src/components/reports/ReportContentEditor.test.tsx
index 042af6fa2..59b6f1b15 100644
--- a/client/src/components/reports/ReportContentEditor.test.tsx
+++ b/client/src/components/reports/ReportContentEditor.test.tsx
@@ -40,6 +40,27 @@
  * `label={content.labels.attachmentsNote}` directly to `EditableField`, which renders a real
  * `<label htmlFor>` association (resolvable via `getByLabelText`), and the old standalone
  * chrome-t `<label>` element that preceded the Usage `EditableField` has been removed entirely.
+ *
+ * #1959 RE-SHAPE. Three changes to this component invalidate the older assertions this file used to
+ * carry, and the tests below now assert the NEW behaviour instead:
+ *
+ *  a. The dedicated **Attachments Note column is gone** (desktop header/cell and mobile card row
+ *     alike). `row.attachmentsNote` is no longer an EditableField at all — it is now READ-ONLY
+ *     text, joined with `row.areaText` by ' · ' into a single grey `.usageMetaText` element that
+ *     sits inside the Usage cell (a `<div>` on desktop, a `<span>` in the mobile card). So there is
+ *     no `row.<id>.attachmentsNote` onFieldChange/onFieldReset wiring left in this component, and
+ *     `content.labels.attachmentsNote` is no longer rendered anywhere. Tests asserting the ABSENCE
+ *     of the old column are therefore paired with a positive assertion that the note text itself
+ *     still renders (inline), so they can't pass by looking at the wrong tree.
+ *
+ *  b. The `†`/`‡` footnote markers are gone. `ReportContentRow.allocatedMarkers` was removed in
+ *     favour of the booleans `isSplit` / `isDepositReduced`, which render as inline grey
+ *     `.inlineNote` spans reading `(content.labels.splitNote)` / `(content.labels.depositReducedNote)`
+ *     appended to the allocated-amount cell — same inline treatment the isDeposit Badge already had.
+ *
+ *  c. Per-column **visibility checkboxes** sit above the table in a `role="group"`, one per
+ *     column, all checked initially; unchecking one removes that column's `<th>`/`<td>` from the
+ *     desktop table AND its mobile-card row. Local state only — no persistence, no callback.
  * The fixture's `labels` values below are deliberately prefixed `REPORT_*_LABEL` — a differently-
  * shaped string from anything the chrome `t` mock would ever echo — so that any test asserting
  * header/caption/source-info/mobile-card-editable-label text is a genuine regression guard: if the
@@ -581,12 +602,19 @@ describe('ReportContentEditor — full field-wiring matrix (onChange + onReset p
     expect(onFieldReset).toHaveBeenCalledWith(key);
   });
 
-  it('calls onFieldChange for the row-level attachmentsNote field', () => {
+  it('#1959: attachmentsNote is READ-ONLY inline text — it renders, but no form control carries it and no onFieldChange fires for it', () => {
     const { onFieldChange, container } = renderEditor({ content: fullContent() });
     const table = getDesktopTable(container);
-    const field = within(table).getByDisplayValue('Note baseline');
-    fireEvent.change(field, { target: { value: 'edited note' } });
-    expect(onFieldChange).toHaveBeenCalledWith('row.inv-1.attachmentsNote', 'edited note');
+    // Positive: the note text IS on screen (so the negatives below are looking at the right tree).
+    expect(within(table).getByText('Note baseline')).toBeInTheDocument();
+    // Negative: it is not the value of any input/textarea — there is nothing to type into.
+    expect(within(table).queryByDisplayValue('Note baseline')).not.toBeInTheDocument();
+    // Editing every remaining editable field in the row never produces an attachmentsNote key.
+    for (const input of within(table).getAllByRole('textbox')) {
+      fireEvent.change(input, { target: { value: 'anything' } });
+    }
+    expect(onFieldChange).toHaveBeenCalled(); // the usage field did fire — the loop was not empty
+    expect(onFieldChange).not.toHaveBeenCalledWith('row.inv-1.attachmentsNote', expect.anything());
   });
 
   it('calls onFieldReset for the row-level usageText field', () => {
@@ -606,12 +634,14 @@ describe('ReportContentEditor — full field-wiring matrix (onChange + onReset p
 });
 
 describe('ReportContentEditor — reset button accessible names (no raw field-identifier leakage)', () => {
-  it('composes the row-level usage/attachmentsNote reset button names from a translated field name, never the raw "usage"/"attachmentsNote" identifier', () => {
+  it('composes the row-level usage reset button name from a translated field name, never the raw "usage" identifier — and #1959 leaves no attachmentsNote reset button at all', () => {
     const content = fullContent();
     const { container } = renderEditor({
       content,
       overrides: {
         'row.inv-1.usageText': content.rows[0]!.usageText,
+        // Still supplied: even with an attachmentsNote override present in the map, the component
+        // must not grow a reset affordance for a field it no longer lets you edit.
         'row.inv-1.attachmentsNote': content.rows[0]!.attachmentsNote as string,
       },
     });
@@ -622,11 +652,14 @@ describe('ReportContentEditor — reset button accessible names (no raw field-id
         name: 'sourceReports.editable.resetFieldAriaLabel::{"field":"sourceReports.table.usage"}',
       }),
     ).toBeInTheDocument();
+
+    // #1959: usage is now the ONLY editable row field, so exactly one reset button per row.
+    expect(within(table).getAllByRole('button', { name: /resetFieldAriaLabel/ })).toHaveLength(1);
     expect(
-      within(table).getByRole('button', {
+      within(table).queryByRole('button', {
         name: 'sourceReports.editable.resetFieldAriaLabel::{"field":"sourceReports.editable.attachmentsNoteLabel"}',
       }),
-    ).toBeInTheDocument();
+    ).not.toBeInTheDocument();
 
     // Neither raw, untranslated identifier ever stands alone as a button's accessible name.
     expect(within(table).queryByRole('button', { name: 'usage' })).not.toBeInTheDocument();
@@ -794,20 +827,50 @@ describe('ReportContentEditor — table rows', () => {
     expect(badge.className).toContain(styles.statusPaid);
   });
 
-  it('renders an Attachments Note column (labeled from content.labels.attachmentsNote) and EditableField only for rows with a non-null note', () => {
+  it('#1959: renders attachmentsNote as grey inline .usageMetaText INSIDE the usage cell — no Attachments Note column header, and only for rows that have a note', () => {
     const rows = [
-      makeRow({ invoiceId: 'inv-1', attachmentsNote: '1 attachment: Invoice' }),
-      makeRow({ invoiceId: 'inv-2', attachmentsNote: null }),
+      // Distinct usageText per row so the cell lookup below is unambiguous.
+      makeRow({
+        invoiceId: 'inv-1',
+        usageText: 'Noted usage',
+        attachmentsNote: '1 attachment: Invoice',
+      }),
+      makeRow({ invoiceId: 'inv-2', usageText: 'Plain usage', attachmentsNote: null }),
     ];
     const { container } = renderEditor({ content: makeContent({ rows }) });
     const table = getDesktopTable(container);
-    expect(within(table).getByText(LABELS.attachmentsNote)).toBeInTheDocument();
-    expect(within(table).getByDisplayValue('1 attachment: Invoice')).toBeInTheDocument();
+
+    // Positive: the note text renders, as a .usageMetaText element, in the SAME <td> as inv-1's
+    // usage input — proving it moved inline rather than merely disappearing.
+    const noteEl = within(table).getByText('1 attachment: Invoice');
+    expect(noteEl.className).toContain(styles.usageMetaText);
+    const usageCell = within(table).getByDisplayValue('Noted usage').closest('td')!;
+    expect(usageCell).toContainElement(noteEl);
+    // inv-2's own usage cell does NOT get the note.
+    expect(within(table).getByDisplayValue('Plain usage').closest('td')!).not.toContainElement(
+      noteEl,
+    );
+
+    // Negative (now safe — we just proved we're looking at the populated tree): the dedicated
+    // column's header label is gone, and the note is not an editable value.
+    expect(within(table).queryByText(LABELS.attachmentsNote)).not.toBeInTheDocument();
+    expect(within(table).queryByDisplayValue('1 attachment: Invoice')).not.toBeInTheDocument();
+
+    // inv-2 (no note) contributes no .usageMetaText element — exactly one exists in the table.
+    expect(table.querySelectorAll(`.${styles.usageMetaText}`)).toHaveLength(1);
   });
 
-  it('omits the Attachments Note column entirely when no row has a non-null note', () => {
-    const rows = [makeRow({ attachmentsNote: null })];
-    renderEditor({ content: makeContent({ rows }) });
+  it('#1959: renders no .usageMetaText element at all when a row has neither attachmentsNote nor areaText', () => {
+    const withMeta = [makeRow({ attachmentsNote: '1 attachment: Invoice' })];
+    const { container: populated } = renderEditor({ content: makeContent({ rows: withMeta }) });
+    // Control: the selector DOES match when meta is present, so the 0-length assertion below is
+    // testing absence of content rather than a typo'd class name.
+    expect(populated.querySelectorAll(`.${styles.usageMetaText}`).length).toBeGreaterThan(0);
+
+    const { container } = renderEditor({
+      content: makeContent({ rows: [makeRow({ attachmentsNote: null, areaText: null })] }),
+    });
+    expect(container.querySelectorAll(`.${styles.usageMetaText}`)).toHaveLength(0);
     expect(screen.queryByText(LABELS.attachmentsNote)).not.toBeInTheDocument();
   });
 
@@ -820,31 +883,36 @@ describe('ReportContentEditor — table rows', () => {
     expect(onFieldChange).toHaveBeenCalledWith('row.inv-42.usageText', 'Changed');
   });
 
-  it('calls onFieldReset with the correct row.<invoiceId>.attachmentsNote key on reset (desktop table)', () => {
-    const rows = [makeRow({ invoiceId: 'inv-42', attachmentsNote: 'Edited note' })];
-    const { onFieldReset, container } = renderEditor({
-      content: makeContent({ rows }),
-      overrides: { 'row.inv-42.attachmentsNote': 'Edited note' },
-    });
-    const table = getDesktopTable(container);
-    const resetButtons = within(table).getAllByRole('button', { name: /resetFieldAriaLabel/ });
-    fireEvent.click(resetButtons[resetButtons.length - 1]!);
-    expect(onFieldReset).toHaveBeenCalledWith('row.inv-42.attachmentsNote');
-  });
-
-  it('calls onFieldReset with the correct row.<invoiceId>.attachmentsNote key on reset (mobile card)', () => {
-    const rows = [makeRow({ invoiceId: 'inv-42', attachmentsNote: 'Edited note' })];
-    const { onFieldReset, container } = renderEditor({
-      content: makeContent({ rows }),
-      overrides: { 'row.inv-42.attachmentsNote': 'Edited note' },
-    });
-    const mobileList = getMobileList(container);
-    const resetButtons = within(mobileList).getAllByRole('button', {
-      name: /resetFieldAriaLabel/,
-    });
-    fireEvent.click(resetButtons[resetButtons.length - 1]!);
-    expect(onFieldReset).toHaveBeenCalledWith('row.inv-42.attachmentsNote');
-  });
+  it.each([
+    ['desktop table', getDesktopTable],
+    ['mobile card', getMobileList],
+  ])(
+    '#1959: clicking every reset button in a row resets ONLY row.<invoiceId>.usageText — attachmentsNote has no reset control (%s)',
+    (_label, getTree) => {
+      const rows = [
+        makeRow({ invoiceId: 'inv-42', usageText: 'Edited usage', attachmentsNote: 'Some note' }),
+      ];
+      const { onFieldReset, container } = renderEditor({
+        content: makeContent({ rows }),
+        // Both keys overridden, so a lingering attachmentsNote affordance would render its reset
+        // button and be caught below.
+        overrides: {
+          'row.inv-42.usageText': 'Edited usage',
+          'row.inv-42.attachmentsNote': 'Some note',
+        },
+      });
+      const tree = within(getTree(container));
+      // Positive: the note text is present in this tree (read-only), so the negative below is not
+      // passing merely because we picked an empty subtree.
+      expect(tree.getByText('Some note')).toBeInTheDocument();
+
+      const resetButtons = tree.getAllByRole('button', { name: /resetFieldAriaLabel/ });
+      expect(resetButtons).toHaveLength(1);
+      for (const btn of resetButtons) fireEvent.click(btn);
+      expect(onFieldReset).toHaveBeenCalledWith('row.inv-42.usageText');
+      expect(onFieldReset).not.toHaveBeenCalledWith('row.inv-42.attachmentsNote');
+    },
+  );
 });
 
 describe('ReportContentEditor — summary rows and footnotes', () => {
@@ -939,8 +1007,238 @@ describe('ReportContentEditor — isDeposit (AC2.1: inline Deposit badge, no mar
   });
 });
 
-describe('ReportContentEditor — areaText (AC5.2/5.3: distinct element below the usage field)', () => {
-  it('renders areaText as a distinct element below the desktop Usage EditableField, not inside its value', () => {
+describe('ReportContentEditor — #1959 isSplit / isDepositReduced inline labels (replacing the † / ‡ footnote markers)', () => {
+  // The allocated cell is a composite of value text + optional badge + optional inline notes, so
+  // assert on the CELL's whole textContent — that is what a user reads — rather than on a single
+  // text node, which would miss ordering/spacing regressions between the runs.
+  function allocatedCellText(container: HTMLElement): string {
+    const table = getDesktopTable(container);
+    // The allocated cell is the only right-aligned <td> holding the €400.00 value; match on the
+    // <td> itself so composite children (badge + inline note spans) are included in textContent.
+    return within(table).getByText(/^€400\.00/, { selector: 'td' }).textContent!;
+  }
+
+  it('appends an inline (splitNote) label to the desktop allocated cell when isSplit, and no † marker', () => {
+    const rows = [
+      makeRow({ invoiceId: 'inv-1', isSplit: true, allocatedAmountValueText: '€400.00' }),
+    ];
+    const { container } = renderEditor({ content: makeContent({ rows }) });
+    expect(allocatedCellText(container)).toBe('€400.00 (REPORT_SPLIT_NOTE_LABEL)');
+    // The label is styled as a grey inline note, not a plain text run.
+    const table = getDesktopTable(container);
+    const noteEl = within(table).getByText('(REPORT_SPLIT_NOTE_LABEL)');
+    expect(noteEl.className).toContain(styles.inlineNote);
+    expect(within(table).queryByText(/†/)).not.toBeInTheDocument();
+  });
+
+  it('appends an inline (depositReducedNote) label to the desktop allocated cell when isDepositReduced, and no ‡ marker', () => {
+    const rows = [
+      makeRow({
+        invoiceId: 'inv-1',
+        isDepositReduced: true,
+        allocatedAmountValueText: '€400.00',
+      }),
+    ];
+    const { container } = renderEditor({ content: makeContent({ rows }) });
+    expect(allocatedCellText(container)).toBe('€400.00 (REPORT_DEPOSIT_REDUCED_NOTE_LABEL)');
+    const table = getDesktopTable(container);
+    expect(within(table).getByText('(REPORT_DEPOSIT_REDUCED_NOTE_LABEL)').className).toContain(
+      styles.inlineNote,
+    );
+    expect(within(table).queryByText(/‡/)).not.toBeInTheDocument();
+  });
+
+  it('renders BOTH inline labels, split before deposit-reduced, when both flags are set', () => {
+    const rows = [
+      makeRow({
+        invoiceId: 'inv-1',
+        isSplit: true,
+        isDepositReduced: true,
+        allocatedAmountValueText: '€400.00',
+      }),
+    ];
+    const { container } = renderEditor({ content: makeContent({ rows }) });
+    expect(allocatedCellText(container)).toBe(
+      '€400.00 (REPORT_SPLIT_NOTE_LABEL) (REPORT_DEPOSIT_REDUCED_NOTE_LABEL)',
+    );
+  });
+
+  it('renders the same inline labels in the mobile card allocated row', () => {
+    const rows = [
+      makeRow({
+        invoiceId: 'inv-1',
+        isSplit: true,
+        isDepositReduced: true,
+        allocatedAmountValueText: '€400.00',
+      }),
+    ];
+    const { container } = renderEditor({ content: makeContent({ rows }) });
+    const card = within(getMobileList(container));
+    expect(card.getByText('(REPORT_SPLIT_NOTE_LABEL)').className).toContain(styles.inlineNote);
+    expect(card.getByText('(REPORT_DEPOSIT_REDUCED_NOTE_LABEL)').className).toContain(
+      styles.inlineNote,
+    );
+  });
+
+  it('renders neither inline label when both flags are false', () => {
+    const rows = [
+      makeRow({
+        invoiceId: 'inv-1',
+        isSplit: false,
+        isDepositReduced: false,
+        allocatedAmountValueText: '€400.00',
+      }),
+    ];
+    const { container } = renderEditor({ content: makeContent({ rows }) });
+    // Positive anchor: the allocated cell rendered its value, so the absences are meaningful.
+    expect(allocatedCellText(container)).toBe('€400.00');
+    expect(screen.queryByText('(REPORT_SPLIT_NOTE_LABEL)')).not.toBeInTheDocument();
+    expect(screen.queryByText('(REPORT_DEPOSIT_REDUCED_NOTE_LABEL)')).not.toBeInTheDocument();
+    expect(container.querySelectorAll(`.${styles.inlineNote}`)).toHaveLength(0);
+  });
+});
+
+describe('ReportContentEditor — #1959 column visibility toggles (local state, no persistence)', () => {
+  function getToggleGroup(container: HTMLElement): HTMLElement {
+    return within(container).getByRole('group', {
+      name: 'sourceReports.editable.columnVisibilityLabel',
+    });
+  }
+
+  it('renders one checked checkbox per column, labeled from content.labels.*, with a Status toggle only for overview reports', () => {
+    const { container } = renderEditor({ content: makeContent({ isOverview: false }) });
+    const group = within(getToggleGroup(container));
+    const boxes = group.getAllByRole('checkbox');
+    expect(boxes).toHaveLength(6); // vendor, invoiceNumber, date, invoiceAmount, allocatedAmount, usage
+    for (const box of boxes) expect(box).toBeChecked();
+    // Labels come from content.labels.* (report language), never a chrome t() echo.
+    for (const label of [
+      LABELS.vendor,
+      LABELS.invoiceNumber,
+      LABELS.date,
+      LABELS.invoiceAmount,
+      LABELS.allocatedAmount,
+      LABELS.usage,
+    ]) {
+      expect(group.getByLabelText(label)).toBeChecked();
+    }
+    expect(group.queryByLabelText(LABELS.status)).not.toBeInTheDocument();
+  });
+
+  it('adds a Status toggle when isOverview is true', () => {
+    const { container } = renderEditor({ content: makeContent({ isOverview: true }) });
+    const group = within(getToggleGroup(container));
+    expect(group.getAllByRole('checkbox')).toHaveLength(7);
+    expect(group.getByLabelText(LABELS.status)).toBeChecked();
+  });
+
+  // Every column gets a DISTINCT value so a disappearing cell can be attributed to the toggled
+  // column and not shadowed by an identical string elsewhere in the table (the fixture's default
+  // invoice/allocated/summary amounts are all €100.00).
+  function distinctValueContent(): ReportContent {
+    return makeContent({
+      rows: [
+        makeRow({
+          invoiceId: 'inv-1',
+          vendor: 'ACME',
+          invoiceNumber: 'INV-001',
+          dateText: '01/10/2026',
+          invoiceAmountText: '€111.00',
+          allocatedAmountValueText: '€222.00',
+          usageText: 'Kitchen work',
+        }),
+      ],
+      summaryRows: [{ key: 'total', label: 'Total', amountText: '€333.00' }],
+    });
+  }
+
+  it.each([
+    [LABELS.vendor, 'ACME'],
+    [LABELS.invoiceNumber, 'INV-001'],
+    [LABELS.date, '01/10/2026'],
+    [LABELS.invoiceAmount, '€111.00'],
+    [LABELS.allocatedAmount, '€222.00'],
+  ])(
+    'unchecking the "%s" toggle removes that column header AND its cell value from the desktop table and the mobile card',
+    (label, cellValue) => {
+      const { container } = renderEditor({ content: distinctValueContent() });
+      const table = getDesktopTable(container);
+      const mobileList = getMobileList(container);
+      const headerCountBefore = table.querySelectorAll('thead th').length;
+
+      // Positive: header label and cell value are both present before the toggle.
+      expect(within(table).getByText(label, { selector: 'th' })).toBeInTheDocument();
+      expect(within(table).getByText(cellValue)).toBeInTheDocument();
+      expect(within(mobileList).getByText(cellValue)).toBeInTheDocument();
+
+      fireEvent.click(within(getToggleGroup(container)).getByLabelText(label));
+
+      // ...and both are gone afterwards, in BOTH responsive trees.
+      expect(within(table).queryByText(label, { selector: 'th' })).not.toBeInTheDocument();
+      expect(within(table).queryByText(cellValue)).not.toBeInTheDocument();
+      expect(within(mobileList).queryByText(cellValue)).not.toBeInTheDocument();
+      expect(table.querySelectorAll('thead th')).toHaveLength(headerCountBefore - 1);
+
+      // The toggle itself stays visible (so the column can be restored) and reflects hidden state.
+      const box = within(getToggleGroup(container)).getByLabelText(label);
+      expect(box).not.toBeChecked();
+      // Re-checking restores the column — the state is a real toggle, not a one-way hide.
+      fireEvent.click(box);
+      expect(within(table).getByText(label, { selector: 'th' })).toBeInTheDocument();
+      expect(within(table).getByText(cellValue)).toBeInTheDocument();
+    },
+  );
+
+  it('unchecking the Usage toggle removes the usage EditableField (and its inline meta text) entirely', () => {
+    const rows = [
+      makeRow({ invoiceId: 'inv-1', usageText: 'Kitchen work', areaText: 'Ground Floor' }),
+    ];
+    const { container } = renderEditor({ content: makeContent({ rows }) });
+    expect(screen.getAllByDisplayValue('Kitchen work').length).toBeGreaterThan(0);
+    expect(container.querySelectorAll(`.${styles.usageMetaText}`).length).toBeGreaterThan(0);
+
+    fireEvent.click(within(getToggleGroup(container)).getByLabelText(LABELS.usage));
+
+    expect(screen.queryAllByDisplayValue('Kitchen work')).toHaveLength(0);
+    expect(container.querySelectorAll(`.${styles.usageMetaText}`)).toHaveLength(0);
+  });
+
+  it('unchecking the Status toggle removes the status Badge from an overview report', () => {
+    const rows = [makeRow({ invoiceId: 'inv-1', status: 'paid', statusText: 'REPORT_PAID_TEXT' })];
+    const { container } = renderEditor({ content: makeContent({ isOverview: true, rows }) });
+    expect(screen.getAllByText('REPORT_PAID_TEXT').length).toBeGreaterThan(0);
+
+    fireEvent.click(within(getToggleGroup(container)).getByLabelText(LABELS.status));
+
+    expect(screen.queryAllByText('REPORT_PAID_TEXT')).toHaveLength(0);
+    expect(
+      within(getDesktopTable(container)).queryByText(LABELS.status, { selector: 'th' }),
+    ).not.toBeInTheDocument();
+  });
+
+  it('hides only the toggled column, leaving the others rendered (toggles are independent)', () => {
+    const { container } = renderEditor();
+    fireEvent.click(within(getToggleGroup(container)).getByLabelText(LABELS.vendor));
+    const table = getDesktopTable(container);
+    expect(within(table).queryByText('ACME')).not.toBeInTheDocument();
+    // Every other column's value survives.
+    expect(within(table).getByText('INV-001')).toBeInTheDocument();
+    expect(within(table).getByText('01/10/2026')).toBeInTheDocument();
+    expect(within(table).getByDisplayValue('Baseline usage')).toBeInTheDocument();
+  });
+
+  it('does not invoke onFieldChange/onFieldReset when a column is toggled (visibility is local state, never an override)', () => {
+    const { container, onFieldChange, onFieldReset } = renderEditor();
+    for (const box of within(getToggleGroup(container)).getAllByRole('checkbox')) {
+      fireEvent.click(box);
+    }
+    expect(onFieldChange).not.toHaveBeenCalled();
+    expect(onFieldReset).not.toHaveBeenCalled();
+  });
+});
+
+describe('ReportContentEditor — areaText / attachmentsNote inline meta (#1959: one grey element inside the usage cell)', () => {
+  it('renders areaText as a distinct .usageMetaText <div> inside the desktop usage cell, not inside the editable value', () => {
     const rows = [
       makeRow({ invoiceId: 'inv-1', usageText: 'Kitchen work', areaText: 'Ground Floor' }),
     ];
@@ -948,27 +1246,77 @@ describe('ReportContentEditor — areaText (AC5.2/5.3: distinct element below th
     const table = getDesktopTable(container);
     const usageInput = within(table).getByDisplayValue('Kitchen work');
     // The area text is not baked into the editable input's value.
-    expect(usageInput).not.toHaveValue('Kitchen work / Ground Floor');
+    expect(usageInput).toHaveValue('Kitchen work');
     const areaEl = within(table).getByText('Ground Floor');
-    expect(areaEl.className).toContain(styles.usageAreaText);
+    expect(areaEl.className).toContain(styles.usageMetaText);
     expect(areaEl.tagName).toBe('DIV');
+    // ...and it lives in the same cell as the usage field (inline, not a separate column).
+    expect(usageInput.closest('td')).toContainElement(areaEl);
   });
 
-  it('renders areaText as a <span> in the mobile card usage row', () => {
+  it('renders areaText as a .usageMetaText <span> in the mobile card usage row', () => {
     const rows = [
       makeRow({ invoiceId: 'inv-1', usageText: 'Kitchen work', areaText: 'Ground Floor' }),
     ];
     const { container } = renderEditor({ content: makeContent({ rows }) });
     const card = within(getMobileList(container));
     const areaEl = card.getByText('Ground Floor');
-    expect(areaEl.className).toContain(styles.usageAreaText);
+    expect(areaEl.className).toContain(styles.usageMetaText);
     expect(areaEl.tagName).toBe('SPAN');
   });
 
-  it('renders no area element (desktop or mobile) when areaText is null', () => {
-    const rows = [makeRow({ invoiceId: 'inv-1', areaText: null })];
+  it.each([
+    ['desktop table', getDesktopTable],
+    ['mobile card', getMobileList],
+  ])(
+    '#1959: joins areaText and attachmentsNote with " · " into ONE meta element, area first (%s)',
+    (_label, getTree) => {
+      const rows = [
+        makeRow({
+          invoiceId: 'inv-1',
+          usageText: 'Kitchen work',
+          areaText: 'Ground Floor',
+          attachmentsNote: '1 attachment: Invoice',
+        }),
+      ];
+      const { container } = renderEditor({ content: makeContent({ rows }) });
+      const tree = getTree(container);
+      // Exact single-node text match pins the join order AND the separator: two sibling elements,
+      // a different separator, or a swapped order would all fail this.
+      const metaEl = within(tree).getByText('Ground Floor · 1 attachment: Invoice');
+      expect(metaEl.className).toContain(styles.usageMetaText);
+      expect(tree.querySelectorAll(`.${styles.usageMetaText}`)).toHaveLength(1);
+    },
+  );
+
+  it('renders areaText alone with no dangling separator when attachmentsNote is null', () => {
+    const rows = [
+      makeRow({
+        invoiceId: 'inv-1',
+        usageText: 'Kitchen work',
+        areaText: 'Ground Floor',
+        attachmentsNote: null,
+      }),
+    ];
     const { container } = renderEditor({ content: makeContent({ rows }) });
-    expect(container.querySelectorAll(`.${styles.usageAreaText}`)).toHaveLength(0);
+    const metaEls = container.querySelectorAll(`.${styles.usageMetaText}`);
+    expect(metaEls).toHaveLength(2); // one desktop <div>, one mobile <span>
+    for (const el of metaEls) {
+      expect(el.textContent).toBe('Ground Floor');
+    }
+  });
+
+  it('renders no meta element (desktop or mobile) when areaText and attachmentsNote are both null', () => {
+    // Control: prove the selector matches when meta IS present, so the 0-length assertion below
+    // cannot pass on a stale/renamed class name.
+    const { container: populated } = renderEditor({
+      content: makeContent({ rows: [makeRow({ areaText: 'Ground Floor' })] }),
+    });
+    expect(populated.querySelectorAll(`.${styles.usageMetaText}`).length).toBeGreaterThan(0);
+
+    const rows = [makeRow({ invoiceId: 'inv-1', areaText: null, attachmentsNote: null })];
+    const { container } = renderEditor({ content: makeContent({ rows }) });
+    expect(container.querySelectorAll(`.${styles.usageMetaText}`)).toHaveLength(0);
   });
 });
 
@@ -1018,7 +1366,7 @@ describe(
       expect(card.queryByText('sourceReports.table.usage')).not.toBeInTheDocument();
     });
 
-    it('renders the mobile card Status label (from content.labels.status) and Badge (labeled from row.statusText) and Attachments Note field (labeled from content.labels.attachmentsNote via a real htmlFor association) consistently with the desktop table', () => {
+    it('renders the mobile card Status label (from content.labels.status) and Badge (labeled from row.statusText), and #1959 renders attachmentsNote as read-only inline meta text with no labeled field of its own', () => {
       const rows = [
         makeRow({
           invoiceId: 'inv-1',
@@ -1031,9 +1379,14 @@ describe(
       const card = within(getMobileList(container));
       expect(card.getByText(LABELS.status)).toBeInTheDocument();
       expect(card.getByText('REPORT_PAID_TEXT')).toBeInTheDocument();
-      const noteField = card.getByLabelText(LABELS.attachmentsNote);
-      expect(noteField).toHaveValue('1 attachment: Invoice');
-      expect(noteField.tagName).toBe('INPUT');
+
+      // Positive: the note text renders inline, as grey meta text.
+      const noteEl = card.getByText('1 attachment: Invoice');
+      expect(noteEl.className).toContain(styles.usageMetaText);
+      expect(noteEl.tagName).toBe('SPAN');
+      // Negative: no <label>-associated form control and no caption for it any more.
+      expect(card.queryByLabelText(LABELS.attachmentsNote)).not.toBeInTheDocument();
+      expect(card.queryByText(LABELS.attachmentsNote)).not.toBeInTheDocument();
       // The old standalone chrome-t label (removed from production code) must never reappear.
       expect(
         card.queryByText('sourceReports.editable.attachmentsNoteLabel'),
@@ -1047,10 +1400,14 @@ describe(
       expect(card.queryByText(LABELS.status)).not.toBeInTheDocument();
     });
 
-    it('omits the mobile card Attachments Note row when the row has no note', () => {
-      const rows = [makeRow({ invoiceId: 'inv-1', attachmentsNote: null })];
+    it('#1959: renders no inline meta text in the mobile card when the row has no note (and the usage field is still there)', () => {
+      const rows = [makeRow({ invoiceId: 'inv-1', attachmentsNote: null, areaText: null })];
       const { container } = renderEditor({ content: makeContent({ rows }) });
-      const card = within(getMobileList(container));
+      const mobileList = getMobileList(container);
+      const card = within(mobileList);
+      // Positive anchor: the card rendered its usage field, so the absences below are meaningful.
+      expect(card.getByLabelText(LABELS.usage)).toHaveValue('Baseline usage');
+      expect(mobileList.querySelectorAll(`.${styles.usageMetaText}`)).toHaveLength(0);
       expect(card.queryByText(LABELS.attachmentsNote)).not.toBeInTheDocument();
       expect(card.queryByLabelText(LABELS.attachmentsNote)).not.toBeInTheDocument();
     });
@@ -1068,7 +1425,7 @@ describe(
       expect(card.getByText('€-200.00 (refund)')).toBeInTheDocument();
     });
 
-    it('wires the mobile card usage/attachmentsNote EditableFields to the same onFieldChange keys as the desktop table', () => {
+    it('wires the mobile card usage EditableField to the same onFieldChange key as the desktop table, and #1959 exposes no attachmentsNote control to wire', () => {
       const rows = [
         makeRow({
           invoiceId: 'inv-7',
@@ -1083,9 +1440,14 @@ describe(
       fireEvent.change(usageInput, { target: { value: 'Changed usage' } });
       expect(onFieldChange).toHaveBeenCalledWith('row.inv-7.usageText', 'Changed usage');
 
-      const noteInput = card.getByDisplayValue('Baseline note');
-      fireEvent.change(noteInput, { target: { value: 'Changed note' } });
-      expect(onFieldChange).toHaveBeenCalledWith('row.inv-7.attachmentsNote', 'Changed note');
+      // Positive: the note text is rendered in this card (read-only)...
+      expect(card.getByText('Baseline note')).toBeInTheDocument();
+      // ...but usage is the card's ONLY editable control, so nothing can emit an attachmentsNote key.
+      expect(card.getAllByRole('textbox')).toEqual([usageInput]);
+      expect(onFieldChange).not.toHaveBeenCalledWith(
+        'row.inv-7.attachmentsNote',
+        expect.anything(),
+      );
     });
 
     it('wires the mobile card usage reset button to the same onFieldReset key as the desktop table', () => {
diff --git a/client/src/lib/reportContent/types.ts b/client/src/lib/reportContent/types.ts
index cc82f2388..ed69aee96 100644
--- a/client/src/lib/reportContent/types.ts
+++ b/client/src/lib/reportContent/types.ts
@@ -19,7 +19,11 @@ export interface ReportContentRow {
   isRefund: boolean;
   refundNoteText: string; // shown only when isRefund
   usageText: string; // EDITABLE — key `row.<invoiceId>.usageText`
-  attachmentsNote: string | null; // EDITABLE when non-null — key `row.<invoiceId>.attachmentsNote`; null = no docs, omitted entirely
+  // READ-ONLY since #1959 moved it inline, with areaText, into the Usage cell's grey meta suffix:
+  // the editor renders both as static text and exposes no input for either. applyOverrides still
+  // honours `row.<invoiceId>.attachmentsNote`, but nothing can produce that key any more — it is
+  // unreachable from the UI. null = no docs, omitted entirely.
+  attachmentsNote: string | null;
   areaText: string | null; // read-only leaf area names, distinct comma-joined
 }
 
diff --git a/client/src/lib/reportPdf/coverLetterPdf.test.ts b/client/src/lib/reportPdf/coverLetterPdf.test.ts
index 52a5ba7d3..9e5b10a62 100644
--- a/client/src/lib/reportPdf/coverLetterPdf.test.ts
+++ b/client/src/lib/reportPdf/coverLetterPdf.test.ts
@@ -51,6 +51,8 @@ function makeContent(overrides: Partial<ReportContent> = {}): ReportContent {
       usage: 'Usage',
       attachmentsNote: 'Attachments Note',
       deposit: 'Deposit',
+      splitNote: 'partial',
+      depositReducedNote: 'less deposit',
       source: 'Source',
       sourceType: 'Source Type',
       reference: 'Reference',
diff --git a/client/src/lib/reportPdf/merge.test.ts b/client/src/lib/reportPdf/merge.test.ts
index 20be712d6..cca87e684 100644
--- a/client/src/lib/reportPdf/merge.test.ts
+++ b/client/src/lib/reportPdf/merge.test.ts
@@ -205,6 +205,8 @@ function makeContent(overrides: Partial<ReportContent> = {}): ReportContent {
       usage: 'Usage',
       attachmentsNote: 'Attachments Note',
       deposit: 'Deposit',
+      splitNote: 'partial',
+      depositReducedNote: 'less deposit',
       source: 'Source',
       sourceType: 'Source Type',
       reference: 'Reference',
diff --git a/client/src/lib/reportPdf/overviewPdf.test.ts b/client/src/lib/reportPdf/overviewPdf.test.ts
index 68e8d0744..e05ed69ac 100644
--- a/client/src/lib/reportPdf/overviewPdf.test.ts
+++ b/client/src/lib/reportPdf/overviewPdf.test.ts
@@ -45,19 +45,18 @@
 import { describe, it, expect } from '@jest/globals';
 import type { TFunction } from 'i18next';
 import type { ReportContent, ReportContentRow } from '../reportContent/index.js';
+import type { UsageCellSegment } from './overviewPdf.js';
 import {
   buildOverviewContent,
   splitIntoPageSafeChunks,
+  packUsageCellRows,
   buildUsageTextRuns,
   USAGE_WIDTH_7COL,
   USAGE_WIDTH_6COL,
   USAGE_SAFE_TOKEN_CHARS_7COL,
   USAGE_SAFE_TOKEN_CHARS_6COL,
   VENDOR_SAFE_TOKEN_CHARS,
-  SMALL_SAFE_TOKEN_CHARS_7COL,
-  SMALL_SAFE_TOKEN_CHARS_6COL,
   MAX_SAFE_USAGE_CHUNK_CHARS,
-  MAX_SAFE_SMALL_CHUNK_CHARS,
 } from './overviewPdf.js';
 import { tableOffsetsTotal, printableWidth } from './pageGeometry.js';
 
@@ -160,6 +159,50 @@ function usageRunsText(text: unknown): string {
   return text as string;
 }
 
+// #1959: `areaText`/`attachmentsNote` are no longer separate stack sub-rows or separate
+// continuation rows — they are appended to the SAME Usage cell as the first `usageText` chunk, as
+// one trailing grey run prefixed with '\n' (see buildOverviewContent's `metaPieces`). Because
+// `buildUsageTextRuns()` tokenizes usageText into one run PER WHITESPACE-DELIMITED TOKEN, the
+// grey meta run is NEVER at a fixed index like `text[1]` — its position depends on the token count
+// of the usage text before it. Split the cell into "the usageText runs" and "the trailing grey
+// meta run (if any)" by the run's own grey color, so assertions address the two by identity rather
+// than by a positional guess that silently drifts with the fixture's word count.
+const GREY = '#6b7280';
+
+function splitUsageCell(cell: unknown): {
+  usageText: string;
+  metaRun: { text: string; color?: string } | null;
+} {
+  const runs = (cell as { text: { text: string; color?: string }[] }).text;
+  if (!Array.isArray(runs)) {
+    throw new Error('Usage cell .text is not a run array — buildUsageTextRuns wiring changed?');
+  }
+  const greyIndexes = runs.map((run, i) => (run.color === GREY ? i : -1)).filter((i) => i !== -1);
+  if (greyIndexes.length > 1) {
+    throw new Error(
+      `Expected at most ONE grey meta run in a Usage cell, found ${greyIndexes.length}`,
+    );
+  }
+  const metaIndex = greyIndexes[0];
+  if (metaIndex === undefined) {
+    return { usageText: runs.map((r) => r.text).join(''), metaRun: null };
+  }
+  // The grey meta run must be the LAST run in the cell — it is a suffix, never interleaved into
+  // the usage prose.
+  if (metaIndex !== runs.length - 1) {
+    throw new Error(
+      `Grey meta run is at index ${metaIndex} of ${runs.length} runs — expected it last`,
+    );
+  }
+  return {
+    usageText: runs
+      .slice(0, metaIndex)
+      .map((r) => r.text)
+      .join(''),
+    metaRun: runs[metaIndex]!,
+  };
+}
+
 function getTable(content: unknown[]): {
   headerRows: number;
   widths: (string | number)[];
@@ -375,6 +418,175 @@ function runsText(runs: { text: string }[]): string {
   return runs.map((run) => run.text).join('');
 }
 
+// ─── packUsageCellRows (#1959 fix round) ──────────────────────────────────────────────────────
+//
+// The bound that makes `dontBreakRows: true` safe now applies to the Usage cell's WHOLE content
+// stream (usage prose + the grey areaText/attachmentsNote suffix), because both render in the same
+// cell at the same 8pt font. Bounding `usageText` alone is what let #1959 silently drop pages: the
+// suffix is unbounded by input (`attachmentsNote` has no maxLength anywhere; `areaText` is
+// aggregate-unbounded across N leaf areas), so an over-tall row was measured and then discarded.
+//
+// These are the packer's own unit tests. Every assertion below is expressed against the INPUT and
+// the declared budget, never against the packer's own output, so a packing regression cannot
+// satisfy them by moving in step.
+describe('packUsageCellRows (#1959 fix round: one page-safe budget for the whole Usage cell)', () => {
+  const BUDGET = 20; // small, so the packing rules are legible in the assertions
+
+  function flatten(rows: UsageCellSegment[][]): UsageCellSegment[] {
+    return rows.flat();
+  }
+  function totalText(rows: UsageCellSegment[][]): string {
+    return flatten(rows)
+      .map((s) => s.text)
+      .join('');
+  }
+  function rowLengths(rows: UsageCellSegment[][]): number[] {
+    return rows.map((row) => row.reduce((sum, s) => sum + s.text.length, 0));
+  }
+  function metaRowIndexes(rows: UsageCellSegment[][]): number[] {
+    return rows.flatMap((row, i) => (row.some((s) => s.meta) ? [i] : []));
+  }
+
+  it('returns a single row, segments unchanged, when the whole stream fits the budget (the dominant case must not grow rows)', () => {
+    const segments: UsageCellSegment[] = [
+      { text: 'Kitchen' },
+      { text: '\nGround Floor', meta: true },
+    ];
+    const rows = packUsageCellRows(segments, BUDGET);
+    expect(rows).toHaveLength(1);
+    expect(rows[0]).toEqual(segments);
+  });
+
+  it('never exceeds the budget on any row, for a stream far larger than one row', () => {
+    const rows = packUsageCellRows(
+      [{ text: 'aaa bbb ccc ddd eee fff ggg hhh iii jjj' }, { text: '\nkkk lll mmm', meta: true }],
+      BUDGET,
+    );
+    expect(rows.length).toBeGreaterThan(1);
+    for (const length of rowLengths(rows)) {
+      expect(length).toBeLessThanOrEqual(BUDGET);
+    }
+  });
+
+  it('is lossless: concatenating every returned segment reproduces the input stream exactly, whitespace included', () => {
+    const prose = 'Materialien und Arbeitsleistung für die Sanierung der Fassade';
+    const meta = '\nErdgeschoss · 2 Anhänge: Angebot, Rechnung';
+    const rows = packUsageCellRows([{ text: prose }, { text: meta, meta: true }], BUDGET);
+    expect(totalText(rows)).toBe(prose + meta);
+  });
+
+  it("preserves each segment's own `meta` flag through splitting — prose pieces never become grey and suffix pieces never become body text", () => {
+    const prose = 'aaa bbb ccc ddd eee fff ggg hhh';
+    const meta = '\niii jjj kkk lll mmm nnn';
+    const rows = packUsageCellRows([{ text: prose }, { text: meta, meta: true }], BUDGET);
+    const flat = flatten(rows);
+    expect(
+      flat
+        .filter((s) => !s.meta)
+        .map((s) => s.text)
+        .join(''),
+    ).toBe(prose);
+    expect(
+      flat
+        .filter((s) => s.meta)
+        .map((s) => s.text)
+        .join(''),
+    ).toBe(meta);
+  });
+
+  it('places the suffix LAST: its final piece always lands on the last row, and no row before the suffix starts carries it', () => {
+    const rows = packUsageCellRows(
+      [{ text: 'aaa bbb ccc ddd eee fff ggg hhh iii' }, { text: '\njjj kkk', meta: true }],
+      BUDGET,
+    );
+    const metaRows = metaRowIndexes(rows);
+    expect(metaRows.length).toBeGreaterThan(0);
+    expect(Math.max(...metaRows)).toBe(rows.length - 1);
+    // Contiguous: once the suffix starts it never yields back to prose rows.
+    expect(metaRows).toEqual(
+      Array.from({ length: metaRows.length }, (_, i) => Math.min(...metaRows) + i),
+    );
+  });
+
+  it('starts a fresh row rather than splitting a suffix that does not fit the leftover but does fit a whole row — keeping the grey run contiguous', () => {
+    // Prose 12 chars leaves 8 of a 20-char budget; the 11-char suffix fits a whole row but not the
+    // leftover, so it must move down intact rather than be sliced across two rows.
+    const rows = packUsageCellRows(
+      [{ text: 'aaaa bbb ccc' }, { text: '\ndddd eeee', meta: true }],
+      BUDGET,
+    );
+    expect(rows).toHaveLength(2);
+    expect(rows[0]!.every((s) => !s.meta)).toBe(true);
+    expect(rows[1]).toEqual([{ text: '\ndddd eeee', meta: true }]);
+  });
+
+  it('splits a suffix genuinely larger than a whole row across as many rows as the budget requires', () => {
+    const meta = `\n${'x'.repeat(BUDGET * 3)}`;
+    const rows = packUsageCellRows([{ text: 'a' }, { text: meta, meta: true }], BUDGET);
+    expect(metaRowIndexes(rows).length).toBeGreaterThan(1);
+    expect(totalText(rows)).toBe('a' + meta);
+    for (const length of rowLengths(rows)) {
+      expect(length).toBeLessThanOrEqual(BUDGET);
+    }
+  });
+
+  it('flushes cleanly when the prose fills a row EXACTLY, putting the suffix on the next row rather than a zero-space row', () => {
+    // 'aaaa bbbb cccc ddddd' is exactly BUDGET (20) characters, so the row is full with zero
+    // leftover when the suffix arrives — the `remaining <= 0` boundary. A regression here would
+    // either emit an empty row or loop.
+    const prose = 'aaaa bbbb cccc ddddd';
+    expect(prose.length).toBe(BUDGET);
+    const rows = packUsageCellRows([{ text: prose }, { text: '\nEG', meta: true }], BUDGET);
+    expect(rows).toHaveLength(2);
+    expect(rows[0]).toEqual([{ text: prose }]);
+    expect(rows[1]).toEqual([{ text: '\nEG', meta: true }]);
+    expect(totalText(rows)).toBe(prose + '\nEG');
+    for (const length of rowLengths(rows)) {
+      expect(length).toBeGreaterThan(0);
+      expect(length).toBeLessThanOrEqual(BUDGET);
+    }
+  });
+
+  it('keeps an empty usageText as its own (empty) leading segment, so the cell still renders a body run', () => {
+    const rows = packUsageCellRows([{ text: '' }, { text: '\nGround Floor', meta: true }], BUDGET);
+    expect(rows).toHaveLength(1);
+    expect(rows[0]![0]).toEqual({ text: '' });
+    expect(rows[0]![1]!.meta).toBe(true);
+  });
+
+  it('returns one empty row for an entirely empty stream, never an empty array (callers index row 0 unconditionally)', () => {
+    expect(packUsageCellRows([], BUDGET)).toEqual([[{ text: '' }]]);
+  });
+
+  it('handles a prose-only stream identically to plain chunking, with no grey segment invented', () => {
+    const prose = 'aaa bbb ccc ddd eee fff ggg hhh iii jjj kkk';
+    const rows = packUsageCellRows([{ text: prose }], BUDGET);
+    expect(totalText(rows)).toBe(prose);
+    expect(flatten(rows).some((s) => s.meta)).toBe(false);
+    expect(rows.map((r) => r[0]!.text)).toEqual(splitIntoPageSafeChunks(prose, BUDGET));
+  });
+
+  it('hard-splits a single whitespace-free token longer than the budget rather than looping forever', () => {
+    const token = 'z'.repeat(BUDGET * 2 + 5);
+    const rows = packUsageCellRows([{ text: token }], BUDGET);
+    expect(totalText(rows)).toBe(token);
+    for (const length of rowLengths(rows)) {
+      expect(length).toBeLessThanOrEqual(BUDGET);
+    }
+  });
+
+  it('at the production budget, a realistic short cell is still exactly one row (guards against a regression that paginates ordinary reports)', () => {
+    const rows = packUsageCellRows(
+      [
+        { text: 'Rohbauarbeiten Erdgeschoss inklusive Bodenplatte' },
+        { text: '\nErdgeschoss · 1 Anhang: Rechnung', meta: true },
+      ],
+      MAX_SAFE_USAGE_CHUNK_CHARS,
+    );
+    expect(rows).toHaveLength(1);
+  });
+});
+
 describe('buildUsageTextRuns (#1929 round 2/3/4 word-break follow-up findings, scenarios 1/2/3)', () => {
   // Pin the actual threshold constants rather than re-typing 16/22 — if pageGeometry.ts's
   // per-char estimate or the USAGE_WIDTH_*COL values ever change, this test's own expectations
@@ -396,13 +608,6 @@ describe('buildUsageTextRuns (#1929 round 2/3/4 word-break follow-up findings, s
     expect(VENDOR_SAFE_TOKEN_CHARS).toBe(Math.floor(45 / (8 * 1.04)));
   });
 
-  it('[#1929 round 4] SMALL_SAFE_TOKEN_CHARS_7COL === 14 and _6COL === 19 (floor(USAGE_WIDTH_*COL / (9 * 1.04)) — the areaText/attachmentsNote continuation-row threshold, at the 9pt "small" style, not the 8pt body font)', () => {
-    expect(SMALL_SAFE_TOKEN_CHARS_7COL).toBe(14);
-    expect(SMALL_SAFE_TOKEN_CHARS_6COL).toBe(19);
-    expect(SMALL_SAFE_TOKEN_CHARS_7COL).toBe(Math.floor(USAGE_WIDTH_7COL / (9 * 1.04)));
-    expect(SMALL_SAFE_TOKEN_CHARS_6COL).toBe(Math.floor(USAGE_WIDTH_6COL / (9 * 1.04)));
-  });
-
   it('[#1929 round 4] MAX_SAFE_USAGE_CHUNK_CHARS === 650, sitting below the measured true ceiling (704) with margin, and comfortably above AC12s 600-char zero-degradation floor', () => {
     // Pin the RELATIONSHIP, not just the literal: 650 must stay strictly between AC12's 600-char
     // floor and the measured true ceiling (704, from a real '№'-only worst-case render — see
@@ -423,15 +628,6 @@ describe('buildUsageTextRuns (#1929 round 2/3/4 word-break follow-up findings, s
     expect(marginFraction).toBeGreaterThanOrEqual(0.07);
   });
 
-  it('[#1929 round 4] MAX_SAFE_SMALL_CHUNK_CHARS === 450, sitting below its own measured true ceiling (546) — the new areaText/attachmentsNote chunking ceiling, no AC-mandated floor unlike usageText', () => {
-    const MEASURED_TRUE_CEILING = 546;
-    expect(MAX_SAFE_SMALL_CHUNK_CHARS).toBe(450);
-    expect(MAX_SAFE_SMALL_CHUNK_CHARS).toBeLessThan(MEASURED_TRUE_CEILING);
-    const marginFraction =
-      (MEASURED_TRUE_CEILING - MAX_SAFE_SMALL_CHUNK_CHARS) / MEASURED_TRUE_CEILING;
-    expect(marginFraction).toBeGreaterThanOrEqual(0.15);
-  });
-
   describe('I1: joining every returned run reconstructs the input exactly', () => {
     it('empty string returns a single empty-text run', () => {
       const runs = buildUsageTextRuns('', USAGE_SAFE_TOKEN_CHARS_7COL);
@@ -771,35 +967,40 @@ describe('buildOverviewContent — row rendering (consumes already-derived Repor
   });
 
   describe('Usage cell: plain text vs inline grey meta text', () => {
-    it('renders a plain { text } cell (not a stack) when both areaText and attachmentsNote are null', () => {
+    it('renders a plain { text } cell (not a stack) with NO grey meta run when both areaText and attachmentsNote are null', () => {
       const row = makeRow({ usageText: 'Kitchen work', areaText: null, attachmentsNote: null });
       const content = makeContent({ rows: [row] });
       const result = buildOverviewContent(content, new Map(), t);
       const table = getTable(result);
       const cell = (table.body[1] as unknown[])[5] as { text?: unknown; stack?: unknown };
       expect(cell.stack).toBeUndefined();
-      expect(usageRunsText(cell.text)).toBe('Kitchen work');
+      const { usageText, metaRun } = splitUsageCell(cell);
+      expect(usageText).toBe('Kitchen work');
+      expect(metaRun).toBeNull();
       // header (1) + 1 usage row + 1 summary row = 3 — no continuation rows.
       expect(table.body).toHaveLength(3);
     });
 
-    it('renders a text array with a grey newline run when attachmentsNote is present', () => {
+    it('#1959: appends attachmentsNote as a trailing grey newline-prefixed run INSIDE the usage cell (not a stack sub-row, not a separate row)', () => {
       const row = makeRow({ usageText: 'Kitchen work', attachmentsNote: '1 attachment: Invoice' });
       const content = makeContent({ rows: [row] });
       const result = buildOverviewContent(content, new Map(), t);
       const table = getTable(result);
-      const cell = (table.body[1] as unknown[])[5] as {
-        text: { text: string; color?: string }[];
-        stack?: unknown;
-      };
+      const cell = (table.body[1] as unknown[])[5] as { text: unknown; stack?: unknown };
       expect(cell.stack).toBeUndefined();
-      expect(Array.isArray(cell.text)).toBe(true);
-      expect(cell.text[0]!.text).toBe('Kitchen work');
-      expect(cell.text[1]!.text).toContain('1 attachment: Invoice');
-      expect(cell.text[1]!.color).toBe('#6b7280');
+      const { usageText, metaRun } = splitUsageCell(cell);
+      // The usage prose itself is untouched — the note is NOT folded into it.
+      expect(usageText).toBe('Kitchen work');
+      expect(usageText).not.toContain('1 attachment: Invoice');
+      // ...and the note IS present, as the grey suffix run, newline-separated from the prose.
+      expect(metaRun).not.toBeNull();
+      expect(metaRun!.text).toBe('\n1 attachment: Invoice');
+      expect(metaRun!.color).toBe(GREY);
+      // The whole thing stays in ONE row — no continuation row is emitted for the note.
+      expect(table.body).toHaveLength(3);
     });
 
-    it('AC5.2: renders areaText and attachmentsNote joined in the grey meta run when both are present', () => {
+    it('#1959 AC5.2: joins areaText and attachmentsNote with " · " in a single trailing grey run, area first', () => {
       const row = makeRow({
         usageText: 'Kitchen work',
         areaText: 'Ground Floor',
@@ -808,16 +1009,17 @@ describe('buildOverviewContent — row rendering (consumes already-derived Repor
       const content = makeContent({ rows: [row] });
       const result = buildOverviewContent(content, new Map(), t);
       const table = getTable(result);
-      const cell = (table.body[1] as unknown[])[5] as {
-        text: { text: string; color?: string }[];
-      };
-      expect(cell.text[0]!.text).toBe('Kitchen work');
-      expect(cell.text[1]!.text).toContain('Ground Floor');
-      expect(cell.text[1]!.text).toContain('1 attachment: Invoice');
-      expect(cell.text[1]!.color).toBe('#6b7280');
+      const cell = (table.body[1] as unknown[])[5] as { text: unknown };
+      const { usageText, metaRun } = splitUsageCell(cell);
+      expect(usageText).toBe('Kitchen work');
+      // Exact string pins the order (area before note), the separator, and the newline prefix —
+      // any of those changing is a visible regression in the rendered PDF.
+      expect(metaRun).not.toBeNull();
+      expect(metaRun!.text).toBe('\nGround Floor · 1 attachment: Invoice');
+      expect(metaRun!.color).toBe(GREY);
     });
 
-    it('renders areaText alone in the grey meta run when attachmentsNote is null', () => {
+    it('#1959: renders areaText alone in the trailing grey run when attachmentsNote is null (no separator, no empty piece)', () => {
       const row = makeRow({
         usageText: 'Kitchen work',
         areaText: 'Ground Floor',
@@ -826,14 +1028,15 @@ describe('buildOverviewContent — row rendering (consumes already-derived Repor
       const content = makeContent({ rows: [row] });
       const result = buildOverviewContent(content, new Map(), t);
       const table = getTable(result);
-      const cell = (table.body[1] as unknown[])[5] as {
-        text: { text: string; color?: string }[];
-        stack?: unknown;
-      };
+      const cell = (table.body[1] as unknown[])[5] as { text: unknown; stack?: unknown };
       expect(cell.stack).toBeUndefined();
-      expect(cell.text[0]!.text).toBe('Kitchen work');
-      expect(cell.text[1]!.text).toContain('Ground Floor');
-      expect(cell.text[1]!.color).toBe('#6b7280');
+      const { usageText, metaRun } = splitUsageCell(cell);
+      expect(usageText).toBe('Kitchen work');
+      expect(metaRun).not.toBeNull();
+      expect(metaRun!.text).toBe('\nGround Floor');
+      // No dangling separator from the absent attachmentsNote.
+      expect(metaRun!.text).not.toContain('·');
+      expect(metaRun!.color).toBe(GREY);
     });
 
     it('[#1929 round 2] the plain-cell Usage text is a run array of the individual whitespace-preserving tokens (buildUsageTextRuns wiring, not a plain string)', () => {
@@ -1030,7 +1233,7 @@ describe('buildOverviewContent — Usage chunking into continuation rows (scenar
     expect(rowTexts(nextRow)[0]).toBe('Next Vendor');
   });
 
-  it('[scenario 11, #1929 round 4 re-shape] areaText/attachmentsNote each render as their OWN continuation row(s), AFTER every usageText chunk row — never stacked into a usage row, each appearing exactly once', () => {
+  it('[scenario 11, #1959 fix round] areaText/attachmentsNote render as ONE inline grey suffix on the LAST usageText row — never mid-prose, never repeated, appearing exactly once', () => {
     const usageText = proseOfLength(MAX_SAFE_USAGE_CHUNK_CHARS * 3);
     const row = makeRow({
       invoiceId: 'inv-1',
@@ -1043,57 +1246,108 @@ describe('buildOverviewContent — Usage chunking into continuation rows (scenar
     const result = buildOverviewContent(content, new Map(), t);
     const table = getTable(result);
 
-    const expectedUsageChunks = splitIntoPageSafeChunks(
-      usageText,
+    // Row count is driven by `packUsageCellRows` over the WHOLE cell stream (prose + suffix), not
+    // by the prose alone — see that function's own unit tests for the packing rules themselves.
+    const expectedRows = packUsageCellRows(
+      [{ text: usageText }, { text: '\nGround Floor · 1 attachment: Invoice', meta: true }],
       MAX_SAFE_USAGE_CHUNK_CHARS,
     ).length;
-    expect(expectedUsageChunks).toBeGreaterThan(1);
-    // areaText/attachmentsNote are short here (well under MAX_SAFE_SMALL_CHUNK_CHARS), so each
-    // contributes exactly one continuation row.
-    const usageRows = table.body.slice(1, 1 + expectedUsageChunks) as unknown[][];
-    const areaRow = table.body[1 + expectedUsageChunks] as { text?: unknown; style?: string }[];
-    const noteRow = table.body[1 + expectedUsageChunks + 1] as {
-      text?: unknown;
-      style?: string;
-    }[];
-
-    // Every usageText chunk row's Usage cell is plain { text } — no stack, no area/attachments
-    // note mixed in (#1929 round 4: these never share a cell with usageText anymore).
-    for (const contRow of usageRows) {
-      const usageCell = contRow[contRow.length - 1] as { text?: unknown; stack?: unknown };
-      expect(usageCell.stack).toBeUndefined();
+    expect(expectedRows).toBeGreaterThan(1);
+    // header (1) + packed rows + summary row (1).
+    expect(table.body).toHaveLength(1 + expectedRows + 1);
+
+    const usageRows = table.body.slice(1, 1 + expectedRows) as unknown[][];
+
+    // PLACEMENT (changed deliberately in the fix round): the suffix trails the prose, so it sits on
+    // the LAST row — not row 0. Pinning it to row 0 rendered grey meta text mid-prose with more
+    // usage text below it, and forced the prose's own chunk boundary to shrink to make room.
+    const lastRow = usageRows[usageRows.length - 1]!;
+    const lastCell = lastRow[lastRow.length - 1] as { text: unknown; stack?: unknown };
+    expect(lastCell.stack).toBeUndefined();
+    const last = splitUsageCell(lastCell);
+    expect(last.metaRun).not.toBeNull();
+    expect(last.metaRun!.text).toBe('\nGround Floor · 1 attachment: Invoice');
+    expect(last.metaRun!.color).toBe(GREY);
+
+    // Every EARLIER row is pure prose — the suffix is neither duplicated nor emitted early.
+    for (const earlierRow of usageRows.slice(0, -1)) {
+      const cell = earlierRow[earlierRow.length - 1] as { text: unknown; stack?: unknown };
+      expect(cell.stack).toBeUndefined();
+      expect(splitUsageCell(cell).metaRun).toBeNull();
     }
 
-    // areaText's OWN row, immediately after every usageText chunk row.
-    const areaCell = areaRow[areaRow.length - 1] as { text: unknown; style?: string };
-    expect(usageRunsText(areaCell.text)).toBe('Ground Floor');
-    expect(areaCell.style).toBe('small');
-    expect(
-      rowTexts(areaRow)
-        .slice(0, areaRow.length - 1)
-        .every((txt) => txt === ''),
-    ).toBe(true);
-
-    // attachmentsNote's OWN row, immediately after areaText's row.
-    const noteCell = noteRow[noteRow.length - 1] as { text: unknown; style?: string };
-    expect(usageRunsText(noteCell.text)).toBe('1 attachment: Invoice');
-    expect(noteCell.style).toBe('small');
-    expect(
-      rowTexts(noteRow)
-        .slice(0, noteRow.length - 1)
-        .every((txt) => txt === ''),
-    ).toBe(true);
-
-    // header (1) + usageChunks + areaText row (1) + attachmentsNote row (1) + summary row (1).
-    expect(table.body).toHaveLength(1 + expectedUsageChunks + 1 + 1 + 1);
-
-    // areaText/attachmentsNote appear EXACTLY once across the WHOLE table (not just this chunked
-    // group) — no duplication, no leakage into a usage row.
-    const allUsageColumnTexts = table.body.map((r) =>
-      usageRunsText((r[r.length - 1] as { text?: unknown })?.text),
-    );
-    expect(allUsageColumnTexts.filter((txt) => txt === 'Ground Floor')).toHaveLength(1);
-    expect(allUsageColumnTexts.filter((txt) => txt === '1 attachment: Invoice')).toHaveLength(1);
+    // I1 (no character is ever lost): concatenating only the usageText PORTION of every row's cell
+    // — i.e. excluding the grey meta suffix — reproduces the original usageText exactly. This is
+    // what proves the inline suffix didn't displace or truncate any prose.
+    const reconstructed = usageRows.map((r) => splitUsageCell(r[r.length - 1]).usageText).join('');
+    expect(reconstructed).toBe(usageText);
+
+    // Every row's Usage cell stays within the ONE page-safe budget — the bound that makes
+    // `dontBreakRows: true` safe. Asserted against the constant, not the packer's own output.
+    for (const usageRow of usageRows) {
+      const { usageText: prose, metaRun } = splitUsageCell(usageRow[usageRow.length - 1]);
+      expect(prose.length + (metaRun?.text.length ?? 0)).toBeLessThanOrEqual(
+        MAX_SAFE_USAGE_CHUNK_CHARS,
+      );
+    }
+
+    // areaText/attachmentsNote text appears EXACTLY once across the WHOLE table — no duplication
+    // onto other rows, no leakage into the summary row.
+    const wholeUsageColumn = table.body
+      .map((r) => usageRunsText((r[r.length - 1] as { text?: unknown })?.text))
+      .join(' ');
+    expect(wholeUsageColumn.split('Ground Floor')).toHaveLength(2);
+    expect(wholeUsageColumn.split('1 attachment: Invoice')).toHaveLength(2);
+  });
+
+  it('[#1959 fix round] when the suffix gets a row of ITS OWN, its leading newline is dropped — the separator must not render as a blank first line in that cell', () => {
+    // 645 chars of prose fits one 650-char row with only 5 to spare, so the 37-char suffix cannot
+    // share it but does fit a row of its own: it becomes the FIRST (and only) run of the next
+    // row's cell. There the '\n' has nothing to separate it from and would render an empty line.
+    const usageText = proseOfLength(645);
+    const row = makeRow({
+      invoiceId: 'inv-1',
+      usageText,
+      areaText: 'Ground Floor',
+      attachmentsNote: '1 attachment: Invoice',
+    });
+    const table = getTable(buildOverviewContent(makeContent({ rows: [row] }), new Map(), t));
+
+    // header (1) + prose row + suffix-only row + summary (1).
+    expect(table.body).toHaveLength(4);
+
+    const proseRow = splitUsageCell((table.body[1] as unknown[])[5]);
+    expect(proseRow.usageText).toBe(usageText);
+    expect(proseRow.metaRun).toBeNull();
+
+    const suffixRow = splitUsageCell((table.body[2] as unknown[])[5]);
+    // Body portion is the empty run buildUsageTextRuns emits for an absent prose segment.
+    expect(suffixRow.usageText).toBe('');
+    expect(suffixRow.metaRun).not.toBeNull();
+    // NO leading newline — this is the whole point of the assertion.
+    expect(suffixRow.metaRun!.text).toBe('Ground Floor · 1 attachment: Invoice');
+    expect(suffixRow.metaRun!.text.startsWith('\n')).toBe(false);
+    expect(suffixRow.metaRun!.color).toBe(GREY);
+  });
+
+  it('[#1959 fix round] a short usage cell whose prose + suffix fit ONE page-safe row still emits exactly one row, suffix inline and last — the dominant case, unchanged by packing', () => {
+    const row = makeRow({
+      invoiceId: 'inv-1',
+      usageText: 'Kitchen work',
+      areaText: 'Ground Floor',
+      attachmentsNote: '1 attachment: Invoice',
+    });
+    const content = makeContent({ rows: [row] });
+    const table = getTable(buildOverviewContent(content, new Map(), t));
+
+    // header (1) + exactly ONE data row + summary (1) — packing must not add rows when the whole
+    // cell already fits, or every ordinary report would grow spurious continuation rows.
+    expect(table.body).toHaveLength(3);
+    const { usageText, metaRun } = splitUsageCell((table.body[1] as unknown[])[5]);
+    expect(usageText).toBe('Kitchen work');
+    expect(metaRun).not.toBeNull();
+    expect(metaRun!.text).toBe('\nGround Floor · 1 attachment: Invoice');
+    expect(metaRun!.color).toBe(GREY);
   });
 });
 
diff --git a/client/src/lib/reportPdf/overviewPdf.ts b/client/src/lib/reportPdf/overviewPdf.ts
index 02aa870d3..0732730f6 100644
--- a/client/src/lib/reportPdf/overviewPdf.ts
+++ b/client/src/lib/reportPdf/overviewPdf.ts
@@ -14,7 +14,6 @@ import {
 import {
   TABLE_BODY_FONT_SIZE,
   TABLE_HEADER_FONT_SIZE,
-  TABLE_SMALL_FONT_SIZE,
   DEFAULT_LINE_HEIGHT,
   usableColumnWidth,
 } from './pageGeometry.js';
@@ -78,26 +77,26 @@ export const USAGE_WIDTH_6COL = usableColumnWidth(6) - USAGE_FIXED_SUM_6COL; //
  * in this table (every column is an explicit numeric width — see USAGE_FIXED_SUM_* above), an
  * under-flagged token can only paint outside its own cell's bounds — it can no longer widen the
  * table itself, which is the failure this constant exists to prevent. (2) MAX_SAFE_USAGE_CHUNK_CHARS
- * and MAX_SAFE_SMALL_CHUNK_CHARS were pinned by DIRECT MEASUREMENT against a real render, not
+ * was pinned by DIRECT MEASUREMENT against a real render, not
  * derived algebraically from this em value — but that measurement used `№` (the widest character
- * the pre-#1939, 124-character scan had found), not `Ѹ`: they sit 7.7% below their measured
- * `№`-fill ceiling of 704 characters, and 17.6% below their measured `№`-fill ceiling of 546
- * characters, respectively (denominator in both cases: that constant's own measured ceiling — see
- * MAX_SAFE_USAGE_CHUNK_CHARS's and MAX_SAFE_SMALL_CHUNK_CHARS's comments). Those percentages are
- * margins against `№`, not against `Ѹ` — re-rendering with a `Ѹ` fill (narrower per-line character
+ * the pre-#1939, 124-character scan had found), not `Ѹ`: it sits 7.7% below its measured
+ * `№`-fill ceiling of 704 characters
+ * (denominator: that constant's own measured ceiling — see
+ * MAX_SAFE_USAGE_CHUNK_CHARS's comment). That percentage is a
+ * margin against `№`, not against `Ѹ` — re-rendering with a `Ѹ` fill (narrower per-line character
  * count, since `Ѹ` is wider) has not been done, so no percentage margin against `Ѹ` is claimed
  * here. Reason (1) covers only horizontal overflow (a token painting outside its own cell); the
- * hazard these two ceilings actually guard against is VERTICAL — an over-tall row silently
+ * hazard this ceiling actually guards against is VERTICAL — an over-tall row silently
  * dropping content under `dontBreakRows` (an I1 violation; see MAX_SAFE_USAGE_CHUNK_CHARS's
  * comment) — and reason (1) says nothing about that. That vertical hazard CAN be derived against
- * `Ѹ` without a new render, because both measured ceilings are exact multiples of their own
- * `№`-fill per-line character count: 704 = 44 lines x 16 chars/line (492.8pt) and 546 = 39 lines x
- * 14 chars/line (491.4pt) — two independent exact multiples landing on the same ~492pt content
- * budget means those measurements pinned a LINE budget (font-size-driven), not a glyph-driven
- * one. Swapping in `Ѹ`'s narrower per-line count (14 chars/line at 8pt, 13 at 9pt) gives derived
- * `Ѹ` ceilings of 44 x 14 = 616 for MAX_SAFE_USAGE_CHUNK_CHARS and 39 x 13 = 507 for
- * MAX_SAFE_SMALL_CHUNK_CHARS. MAX_SAFE_SMALL_CHUNK_CHARS (450) is safe under this derivation
- * (11.2% under 507); MAX_SAFE_USAGE_CHUNK_CHARS (650) is NOT — it is 34 characters (3 lines,
+ * `Ѹ` without a new render, because the measured ceiling is an exact multiple of its own
+ * `№`-fill per-line character count: 704 = 44 lines x 16 chars/line (492.8pt). (#1929 round 4's
+ * since-removed 9pt companion ceiling measured 546 = 39 lines x 14 chars/line (491.4pt) — two
+ * independent exact multiples landing on the same ~492pt content
+ * budget, which is what showed these measurements pinned a LINE budget (font-size-driven), not a
+ * glyph-driven one.) Swapping in `Ѹ`'s narrower per-line count (14 chars/line at 8pt) gives a derived
+ * `Ѹ` ceiling of 44 x 14 = 616 for MAX_SAFE_USAGE_CHUNK_CHARS.
+ * MAX_SAFE_USAGE_CHUNK_CHARS (650) does NOT clear it — it is 34 characters (3 lines,
  * 33.6pt) over its derived 616-character `Ѹ` ceiling. That overage is tolerated on input
  * reachability, not on either reason above: it requires a 650-character unbroken run of archaic
  * Church Slavonic Uk in a single Usage cell, the same non-credible-input class as `markerText`'s
@@ -123,7 +122,6 @@ export const USAGE_WIDTH_6COL = usableColumnWidth(6) - USAGE_FIXED_SUM_6COL; //
 const WORST_CASE_CHAR_ADVANCE_EM = 1.04;
 const BODY_WORST_CASE_CHAR_WIDTH_PT = TABLE_BODY_FONT_SIZE * WORST_CASE_CHAR_ADVANCE_EM; // 8.32pt
 const HEADER_WORST_CASE_CHAR_WIDTH_PT = TABLE_HEADER_FONT_SIZE * WORST_CASE_CHAR_ADVANCE_EM; // 10.4pt
-const SMALL_WORST_CASE_CHAR_WIDTH_PT = TABLE_SMALL_FONT_SIZE * WORST_CASE_CHAR_ADVANCE_EM; // 9.36pt
 
 /**
  * A single Usage cell whose `usageText` exceeds this length is split into multiple table rows
@@ -133,19 +131,29 @@ const SMALL_WORST_CASE_CHAR_WIDTH_PT = TABLE_SMALL_FONT_SIZE * WORST_CASE_CHAR_A
  * "no character is ever lost" outranks I3 "each row on one page"), a row that genuinely cannot
  * fit one page must still render completely, even if that means it spans pages.
  *
- * This bounds ONLY `usageText`'s own contribution to a row's height — `areaText` and
- * `attachmentsNote` are rendered on their OWN separate, independently-chunked continuation
- * row(s) (see MAX_SAFE_SMALL_CHUNK_CHARS and buildOverviewContent's row-building loop), never
- * stacked into the same cell as a usageText chunk. #1929 round-4 architect review HIGH: the
- * round-3 design stacked areaText/attachmentsNote onto the LAST usageText chunk's row, leaving
- * their COMBINED height unbounded — attachmentsNote has no maxLength anywhere (editor or
+ * SCOPE: this bounds the Usage cell's ENTIRE content stream — `usageText` plus the grey
+ * `areaText`/`attachmentsNote` suffix that #1959 renders inline in the same cell — not
+ * `usageText` alone. The quantity that has to be bounded is "total height of everything in a
+ * row's Usage cell"; see packUsageCellRows, which packs that whole stream against this constant.
+ * Both parts render at the same font ('tableCell', 8pt — the suffix only overrides `color`), so
+ * ONE character budget governs the whole cell. (#1929 round 4's separate 9pt ceiling for the
+ * suffix's own continuation rows was removed with those rows: a 9pt budget is simply the wrong
+ * bound for 8pt content, so it must not be reinstated as-is if a 9pt cell style ever returns.)
+ *
+ * Getting this scope wrong has now cost two rounds, in both directions. #1929 round 3 stacked
+ * areaText/attachmentsNote onto the LAST usageText chunk's row while bounding only usageText,
+ * leaving their COMBINED height unbounded — attachmentsNote has no maxLength anywhere (editor or
  * server) and areaText is aggregate-unbounded (N leaf areas x 200 chars each); measured
  * combinations (700-char usageText + 400-char attachmentsNote = 665.8pt; 700-char usageText +
  * 20-leaf-area areaText = 691.0pt; 2000-char attachmentsNote alone = 1119.4pt) all exceeded the
  * 634.89pt page budget and were silently dropped (rows requiring 3 and 9 pages both rendered as
- * 2). Splitting each field onto its own row means every row's Usage cell now holds AT MOST ONE
- * bounded chunk of ONE field — the quantity actually bounded is "total height of everything in
- * a row's Usage cell", not just usageText's length.
+ * 2). Round 4 fixed that by giving each field independently-chunked continuation rows of its
+ * OWN. #1959 then moved both fields back inline as a single unchunked grey run — reinstating
+ * round 3's defect verbatim (measured on that build: 16,000 characters of attachmentsNote
+ * rendered as 2 pages, the same 16,000 as chunked usageText needs 9; page count SATURATED at 2
+ * and even went 3 -> 2 as content grew, while rendered line count kept growing linearly).
+ * The bound now follows the rendered cell rather than a single field, so it cannot be detached
+ * from #1959's inline design by a future layout change the way round 4's per-field rows were.
  *
  * MEASURED CEILING (AC12, #1929 round-4 architect review): measured directly against a real
  * render of the production table shape (7-column, `headerRows: 1` with word-break-protected
@@ -163,25 +171,6 @@ const SMALL_WORST_CASE_CHAR_WIDTH_PT = TABLE_SMALL_FONT_SIZE * WORST_CASE_CHAR_A
  */
 export const MAX_SAFE_USAGE_CHUNK_CHARS = 650;
 
-/**
- * An `areaText` or `attachmentsNote` value exceeding this length is split into multiple
- * continuation rows the same way an over-long `usageText` is (see MAX_SAFE_USAGE_CHUNK_CHARS).
- * Rendered at 'small' style (9pt, taller line height and wider worst-case glyphs than 'tableCell'
- * at 8pt), so it needs its OWN ceiling, not a reuse of MAX_SAFE_USAGE_CHUNK_CHARS (#1929 round-4
- * architect review HIGH — see that constant's comment for why these two fields must never share
- * a row with usageText, or with each other, in the first place).
- *
- * MEASURED CEILING: same method and production shape as MAX_SAFE_USAGE_CHUNK_CHARS, with the
- * continuation row rendered at 'small' style and the same pre-#1939 worst-case '№' fill (see that
- * constant's comment for why '№', not `Ѹ`, is the character this ceiling was measured against).
- * Measured ceiling: 546 characters. This constant sits ~17.6% below that measured '№'-fill
- * ceiling (denominator: 546) — no AC mandates a specific floor for these two fields (unlike
- * usageText's AC12), so a larger margin is used than MAX_SAFE_USAGE_CHUNK_CHARS's tighter [600,
- * ceiling) window allows. As with that constant, this 17.6% is a margin against `№`; it has not
- * been re-measured against `Ѹ`.
- */
-export const MAX_SAFE_SMALL_CHUNK_CHARS = 450;
-
 /**
  * Splits `text` into chunks no longer than `maxChars`, breaking only at whitespace boundaries
  * except when a single whitespace-free run itself exceeds maxChars, in which case it is
@@ -215,6 +204,103 @@ export function splitIntoPageSafeChunks(text: string, maxChars: number): string[
   return chunks;
 }
 
+/**
+ * One styled piece of a Usage cell's content stream. `meta: true` marks the grey
+ * areaText/attachmentsNote suffix (#1959's inline meta design); everything else renders in the
+ * body colour. Both render at the SAME font ('tableCell', 8pt) — the suffix only overrides
+ * `color` — which is why ONE character budget (MAX_SAFE_USAGE_CHUNK_CHARS) governs the whole
+ * stream, rather than a separate ceiling per styled piece.
+ */
+export interface UsageCellSegment {
+  text: string;
+  meta?: boolean;
+}
+
+/**
+ * Packs a Usage cell's FULL content stream (usage prose, then the grey meta suffix) into per-row
+ * segment groups, each holding at most `maxChars` characters — the bound that makes every emitted
+ * row's Usage cell page-safe under `dontBreakRows: true`.
+ *
+ * This is the round-4 bound reinstated at the right scope. #1929 round 4 bounded the same quantity
+ * by giving `areaText`/`attachmentsNote` continuation rows of their OWN; #1959 moved them inline
+ * into the usage cell as a grey suffix, which restored the unbounded-cell failure mode round 4
+ * fixed (pdfmake measures an over-tall row, then `PageElementWriter` silently DROPS whatever does
+ * not fit — page count saturates at 2 while rendered line count keeps growing). The bound must
+ * therefore apply to what the cell actually renders, which is now usage + area + attachments
+ * TOGETHER, not to `usageText` alone.
+ *
+ * Packing is greedy, so #1959's visual intent is preserved exactly wherever it can be: whenever
+ * the whole cell fits one page-safe row (the overwhelmingly common case), the output is a single
+ * row whose runs are the usage prose followed by the `\n`-prefixed grey suffix — byte-identical to
+ * the pre-fix output. The suffix only lands on a continuation row when the combined content
+ * genuinely cannot fit one row, and it is always emitted LAST in the stream, so it stays the
+ * trailing grey run of the last row it appears on (never split across two rows of one group more
+ * than the budget forces, and never more than ONE grey run per cell).
+ *
+ * Lossless (#1929 I1): concatenating every returned segment's `text`, in order, reconstructs the
+ * input stream exactly — no character, including whitespace, is added or dropped. Empty segments
+ * are preserved so an empty `usageText` still renders its own (empty) leading run.
+ */
+export function packUsageCellRows(
+  segments: UsageCellSegment[],
+  maxChars: number,
+): UsageCellSegment[][] {
+  const rows: UsageCellSegment[][] = [];
+  let current: UsageCellSegment[] = [];
+  let used = 0;
+
+  const flush = (): void => {
+    if (current.length > 0) {
+      rows.push(current);
+      current = [];
+      used = 0;
+    }
+  };
+
+  for (const segment of segments) {
+    if (segment.text.length === 0) {
+      // Consumes no budget; kept so an empty usageText still contributes its own run.
+      current.push({ ...segment });
+      continue;
+    }
+    let rest = segment.text;
+    while (rest.length > 0) {
+      const remaining = maxChars - used;
+      if (remaining <= 0) {
+        flush();
+        continue;
+      }
+      if (rest.length <= remaining) {
+        current.push({ ...segment, text: rest });
+        used += rest.length;
+        break;
+      }
+      if (used > 0 && rest.length <= maxChars) {
+        // Doesn't fit this row's leftover space, but fits a row of its own: start a fresh row
+        // rather than splitting it into a sliver. This keeps the grey meta suffix ONE contiguous
+        // run whenever it fits a page-safe row at all.
+        flush();
+        continue;
+      }
+      if (used > 0) {
+        // Genuinely larger than a whole row — fill this row's leftover space, then carry on.
+        const [head, ...tail] = splitIntoPageSafeChunks(rest, remaining);
+        current.push({ ...segment, text: head! });
+        rest = tail.join('');
+        flush();
+        continue;
+      }
+      const [head, ...tail] = splitIntoPageSafeChunks(rest, maxChars);
+      current.push({ ...segment, text: head! });
+      rest = tail.join('');
+      flush();
+    }
+  }
+  flush();
+
+  return rows.length > 0 ? rows : [[{ text: '' }]];
+}
+
 /**
  * A whitespace-free run at or under this many characters is guaranteed to fit on one line within
  * the given column width, EVEN IN THE WORST CASE (every character as wide as 'W'), and renders
@@ -258,21 +344,6 @@ export const USAGE_SAFE_TOKEN_CHARS_6COL = safeTokenChars(
  */
 export const VENDOR_SAFE_TOKEN_CHARS = safeTokenChars(VENDOR_WIDTH, BODY_WORST_CASE_CHAR_WIDTH_PT);
 
-/**
- * Safe-token-char threshold for the Usage column's areaText/attachmentsNote continuation rows,
- * rendered at 'small' style (9pt) — needs its own threshold rather than reusing
- * USAGE_SAFE_TOKEN_CHARS_*COL, which is derived from the body font's (8pt) worst-case glyph
- * width, narrower than 'small' (#1929 round-4 architect review HIGH).
- */
-export const SMALL_SAFE_TOKEN_CHARS_7COL = safeTokenChars(
-  USAGE_WIDTH_7COL,
-  SMALL_WORST_CASE_CHAR_WIDTH_PT,
-);
-export const SMALL_SAFE_TOKEN_CHARS_6COL = safeTokenChars(
-  USAGE_WIDTH_6COL,
-  SMALL_WORST_CASE_CHAR_WIDTH_PT,
-);
-
 /**
  * Splits `text` into inline pdfmake text runs. Whitespace-free runs at or under `safeTokenChars`
  * are emitted verbatim (default whitespace-only wrapping); a run over that length gets
@@ -374,8 +445,13 @@ export const HEADER_ROW_HEIGHT_MAX =
  * - `refundNoteText`    — fixed `reportT` string (bounded by construction, same as above).
  * - `allocatedAmountValueText` — via `reportFormatters.formatCurrency` (bounded by construction:
  *                          a formatted number, not user input).
- * - `usageText` / `areaText` / `attachmentsNote` — chunked via splitIntoPageSafeChunks against
- *                          MAX_SAFE_USAGE_CHUNK_CHARS / MAX_SAFE_SMALL_CHUNK_CHARS.
+ * - `usageText` / `areaText` / `attachmentsNote` — all three share ONE Usage cell (#1959 renders
+ *                          areaText/attachmentsNote as an inline grey suffix), and that cell's
+ *                          whole content stream is packed into page-safe rows by
+ *                          packUsageCellRows against MAX_SAFE_USAGE_CHUNK_CHARS. VERTICAL height
+ *                          is therefore bounded for all three together. The grey suffix is not
+ *                          routed through buildUsageTextRuns, so HORIZONTALLY it is in the same
+ *                          recorded-not-fixed class as `invoiceNumber` below.
  * - `markerText`        — UNBOUNDED. One `*N` footnote marker is appended per skipped document on
  *                          an invoice (see the row-building loop's `skipMarkers` accumulation),
  *                          with no chunking and no word-break, into the 75pt ALLOCATED_AMOUNT_WIDTH
@@ -386,10 +462,17 @@ export const HEADER_ROW_HEIGHT_MAX =
  *                          self-hosted app — not a credible input, and the fix (chunking a
  *                          footnote-marker run) would be pure ceremony against that reachability.
  *
- * Two channels are recorded here without a fix, same class as `markerText` above:
+ * Three channels are recorded here without a fix, same class as `markerText` above:
  * - `invoiceNumber` (see above) does not route through buildUsageTextRuns, so a 100-character
  *   unbroken number would paint outside its 63pt INVOICE_NUMBER_WIDTH column. Interior column,
  *   cosmetic overflow only, capped at 100 by the server schema — recorded, not fixed.
+ * - the inline grey `areaText`/`attachmentsNote` suffix (see above) is emitted as ONE run rather
+ *   than per-token runs, so an unbroken token in a leaf-area name or note wider than the Usage
+ *   column paints outside that cell. Horizontal and cosmetic only (every column is an explicit
+ *   numeric width, so no token can widen the table itself — see WORST_CASE_CHAR_ADVANCE_EM reason
+ *   (1)); the VERTICAL channel, which is the one that loses content, is closed. Emitting per-token
+ *   runs here would also put more than one grey run in a cell, which readers of these cells
+ *   currently treat as an invariant violation — recorded, not fixed.
  * - `markerText` (see above) is the one remaining unbounded row-height contributor in this table.
  */
 
@@ -582,6 +665,32 @@ export function buildOverviewContent(
     ? USAGE_SAFE_TOKEN_CHARS_7COL
     : USAGE_SAFE_TOKEN_CHARS_6COL;
 
+  /**
+   * Renders one packed row's worth of Usage-cell segments (see packUsageCellRows) into a cell.
+   *
+   * Body segments go through `buildUsageTextRuns` for per-token break-all protection; the grey
+   * meta suffix is emitted as ONE run, so a cell never holds more than a single grey run and that
+   * run is always last (both properties are relied on when reading these cells back). The meta
+   * suffix is NOT token-protected: it is 8pt body-font text in a fixed-width column, so an
+   * over-wide unbroken token there paints outside its own cell — cosmetic horizontal overflow only,
+   * the same recorded-not-fixed class as `invoiceNumber` (see the channel enumeration above).
+   */
+  function buildUsageCell(segments: UsageCellSegment[]): Content {
+    const runs: Content[] = [];
+    segments.forEach((segment, index) => {
+      if (!segment.meta) {
+        runs.push(...buildUsageTextRuns(segment.text, usageSafeTokenChars));
+        return;
+      }
+      // The '\n' separates the suffix from the usage prose it follows. When the suffix STARTS a
+      // cell (it was pushed onto a continuation row of its own), that newline would render an
+      // empty first line instead, so it is dropped — a presentational separator, not content.
+      const text = index === 0 ? segment.text.replace(/^\n/, '') : segment.text;
+      runs.push({ text, color: DEPOSIT_NOTE_TEXT_COLOR });
+    });
+    return { text: runs, style: 'tableCell' };
+  }
+
   for (const contentRow of reportContent.rows) {
     // Allocated amount with skip markers and inline labels
     const skipMarkers = skipFootnotesByInvoiceId.get(contentRow.invoiceId) ?? [];
@@ -619,37 +728,38 @@ export function buildOverviewContent(
       allocatedRuns.push({ text: ` ${contentRow.refundNoteText}` });
     }
 
-    // Usage text is split into page-safe chunks (#1929 AC2/AC4/AC12/HIGH 4): a single Usage
-    // cell that is too long to safely fit one page's printable height is rendered as multiple
-    // table rows instead of one unbreakable (and potentially content-dropping) row. The FIRST
-    // chunk shares this invoice's leading/amount-cell row; any further chunks (rare-by-
-    // construction: AC12 requires 600 chars with zero degradation, well under
-    // MAX_SAFE_USAGE_CHUNK_CHARS) get their own continuation row — no "continued" marker,
-    // per the product-owner's explicit ruling.
-    const usageChunks = splitIntoPageSafeChunks(contentRow.usageText, MAX_SAFE_USAGE_CHUNK_CHARS);
-    // Area and attachmentsNote appear inline in grey after the first usage chunk
+    // Area and attachmentsNote render inline as one trailing grey suffix on the Usage cell
+    // (#1959), '\n'-prefixed and joined by ' · '.
     const metaPieces: string[] = [];
     if (contentRow.areaText) metaPieces.push(contentRow.areaText);
     if (contentRow.attachmentsNote) metaPieces.push(contentRow.attachmentsNote);
-    const firstUsageRuns = buildUsageTextRuns(usageChunks[0]!, usageSafeTokenChars);
+
+    // The Usage cell's ENTIRE content stream — usage prose plus that grey suffix — is packed into
+    // page-safe rows against ONE budget. Bounding `usageText` alone is not enough: the suffix is
+    // rendered in the same cell, at the same font, and neither `attachmentsNote` (no maxLength
+    // anywhere) nor `areaText` (aggregate-unbounded across N leaf areas) is bounded by input, so
+    // an unbounded suffix makes the row over-tall and `dontBreakRows: true` silently DROPS the
+    // overflow instead of paginating it (#1929 architect review HIGH 4 / I1). Wherever the whole
+    // cell fits one row — the common case — this emits exactly one row, unchanged from #1959.
+    // The first packed row shares this invoice's leading/amount cells; any further row is a
+    // Usage-only continuation row with no "continued" marker, per the product-owner's ruling
+    // (#1929 AC2/AC4/AC12).
+    const cellSegments: UsageCellSegment[] = [{ text: contentRow.usageText }];
     if (metaPieces.length > 0) {
-      firstUsageRuns.push({ text: '\n' + metaPieces.join(' · '), color: DEPOSIT_NOTE_TEXT_COLOR });
+      cellSegments.push({ text: `\n${metaPieces.join(' · ')}`, meta: true });
     }
-    const firstUsageCell: Content = { text: firstUsageRuns, style: 'tableCell' };
+    const packedCellRows = packUsageCellRows(cellSegments, MAX_SAFE_USAGE_CHUNK_CHARS);
+
     rows.push([
       ...buildLeadingCells(contentRow, reportContent.isOverview, contentRow.statusText ?? ''),
       ...buildAmountCells(contentRow, allocatedRuns),
-      firstUsageCell,
+      buildUsageCell(packedCellRows[0]!),
     ]);
-    for (let i = 1; i < usageChunks.length; i++) {
-      const cell: Content = {
-        text: buildUsageTextRuns(usageChunks[i]!, usageSafeTokenChars),
-        style: 'tableCell',
-      };
+    for (let i = 1; i < packedCellRows.length; i++) {
       rows.push([
         ...buildEmptyLeadingCells(reportContent.isOverview),
         ...buildEmptyAmountCells(),
-        cell,
+        buildUsageCell(packedCellRows[i]!),
       ]);
     }
   }
diff --git a/client/src/lib/reportPdf/pageGeometry.test.ts b/client/src/lib/reportPdf/pageGeometry.test.ts
index dfca96d76..09459aecb 100644
--- a/client/src/lib/reportPdf/pageGeometry.test.ts
+++ b/client/src/lib/reportPdf/pageGeometry.test.ts
@@ -27,6 +27,7 @@ import {
   tableOffsetsTotal,
   usableColumnWidth,
   headerFootprint,
+  PDF_STYLES,
 } from './pageGeometry.js';
 
 describe('pageGeometry — page constants', () => {
@@ -56,8 +57,14 @@ describe('pageGeometry — page constants', () => {
     expect(TABLE_HEADER_FONT_SIZE).toBe(10);
   });
 
-  it('[#1929 round 4] TABLE_SMALL_FONT_SIZE is 9, consumed by PDF_STYLES.small below (#1939 relocated PDF_STYLES into this module) so overviewPdf.ts areaText/attachmentsNote continuation-row word-break threshold (SMALL_SAFE_TOKEN_CHARS_*COL) is computed from the same constant PDF_STYLES renders with', () => {
+  it('TABLE_SMALL_FONT_SIZE is 9 and is the size PDF_STYLES.small actually renders with (#1939 relocated PDF_STYLES into this module, so the constant and the style it feeds cannot drift apart)', () => {
     expect(TABLE_SMALL_FONT_SIZE).toBe(9);
+    // The single-source-of-truth relationship, not just the literal: the rendered style reads
+    // this constant. #1959's fix round removed the Usage-column 9pt consumers this test used to
+    // reference (areaText/attachmentsNote continuation rows) — the remaining consumers are the
+    // report's secondary text blocks, which still render through PDF_STYLES.small.
+    expect(PDF_STYLES['small']).toBeDefined();
+    expect(PDF_STYLES['small']!.fontSize).toBe(TABLE_SMALL_FONT_SIZE);
   });
 
   it('DEFAULT_LINE_HEIGHT matches merge.ts defaultStyle.lineHeight (1.4)', () => {
diff --git a/client/src/lib/reportPdf/pageGeometry.ts b/client/src/lib/reportPdf/pageGeometry.ts
index e804534f4..5ef40606e 100644
--- a/client/src/lib/reportPdf/pageGeometry.ts
+++ b/client/src/lib/reportPdf/pageGeometry.ts
@@ -45,10 +45,14 @@ export const TABLE_BODY_FONT_SIZE = 8;
 export const TABLE_HEADER_FONT_SIZE = 10;
 
 /**
- * Font size, pt, for the 'small' style used by the Usage column's areaText/attachmentsNote
- * continuation rows (#1929 round-4 architect review HIGH — see overviewPdf.ts's
- * MAX_SAFE_SMALL_CHUNK_CHARS). Consumed by `PDF_STYLES.small` below for the same
- * single-source-of-truth reason as TABLE_HEADER_FONT_SIZE above.
+ * Font size, pt, for the 'small' style. Consumed by `PDF_STYLES.small` below for the same
+ * single-source-of-truth reason as TABLE_HEADER_FONT_SIZE above. Its live consumers are the
+ * report's secondary text: source-info/footnote blocks (overviewPdf.ts), the cover letter's
+ * date/reference lines (coverLetterPdf.ts), and the running page header/footer (shared.ts).
+ *
+ * NOT the Usage column: #1929 round 4 briefly rendered areaText/attachmentsNote continuation rows
+ * at this size, but #1959 moved that content inline into the Usage cell at TABLE_BODY_FONT_SIZE
+ * (8pt), and the fix round bounded it with the usage budget. Nothing in the Usage column is 9pt.
  */
 export const TABLE_SMALL_FONT_SIZE = 9;
 
diff --git a/client/src/lib/reportPdf/realRender.test.ts b/client/src/lib/reportPdf/realRender.test.ts
index 9b7968262..b340f9dc5 100644
--- a/client/src/lib/reportPdf/realRender.test.ts
+++ b/client/src/lib/reportPdf/realRender.test.ts
@@ -78,10 +78,9 @@ import {
   USAGE_WIDTH_7COL,
   USAGE_WIDTH_6COL,
   MAX_SAFE_USAGE_CHUNK_CHARS,
-  MAX_SAFE_SMALL_CHUNK_CHARS,
   HEADER_ROW_HEIGHT_MAX,
-  splitIntoPageSafeChunks,
 } from './overviewPdf.js';
+import type { UsageCellSegment } from './overviewPdf.js';
 
 // ─── Real i18next instance (NOT the app singleton — no localStorage/navigator dependency) ─────
 
@@ -131,11 +130,14 @@ beforeAll(async () => {
 // Crucially, `pdfContent` is passed BY REFERENCE and mutated in place by pdfmake during
 // `getBlob()` — callers keep their own reference to `pdfContent` and read `._calcWidth`/
 // `.positions` off it afterwards.
+// Returns the rendered Blob so callers that need the REAL page count (see the cell-scope
+// invariant block) can load it with pdf-lib without rendering the same tree twice; callers that
+// only need the by-reference layout mutations can ignore the return value.
 async function renderOverviewPdfContent(
   pdfContent: Content[],
   header: { tableTitle: string; sourceName: string },
   t: TFunction,
-): Promise<void> {
+): Promise<Blob> {
   const { pdfMake } = await loadPdfLibs();
   const pdfDoc = pdfMake.createPdf({
     content: pdfContent,
@@ -153,7 +155,7 @@ async function renderOverviewPdfContent(
     defaultStyle: PDF_DEFAULT_STYLE,
     styles: PDF_STYLES,
   });
-  await pdfDoc.getBlob();
+  return pdfDoc.getBlob();
 }
 
 // #1932 (AC 1.2/7.2): renders a cover-letter Content[] tree (no table/header/footer involved,
@@ -279,6 +281,73 @@ function usageCellText(text: unknown): string {
   return text as string;
 }
 
+// #1959: `areaText`/`attachmentsNote` are no longer separate continuation rows — they are appended
+// to the SAME Usage cell as the usage prose, as a trailing grey run prefixed with '\n'. Its index
+// within the run array depends on the token count of the usage prose before it (see
+// buildUsageTextRuns), so it must be located by its own grey color, never by a fixed position.
+//
+// #1959 fix round: `packUsageCellRows` packs the cell's WHOLE content stream (prose + grey suffix)
+// against one page-safe budget, so a long suffix now spans SEVERAL rows — one grey run per row,
+// always the last run of its own cell. Reconstructing the suffix therefore means concatenating the
+// grey run of EVERY row in the group, not reading a single cell (see `readCellGroup` below); a
+// helper that only looked at row 0 was what let the old page-count tripwire fail for the wrong
+// reason instead of flipping.
+const META_GREY = '#6b7280';
+
+function splitUsageCell(cell: unknown): {
+  usageText: string;
+  /** Grey suffix exactly as rendered, INCLUDING any leading '\n' — for faithful concatenation. */
+  metaRaw: string | null;
+  /** Grey suffix with the presentational leading '\n' stripped — for single-cell assertions. */
+  metaText: string | null;
+} {
+  const runs = (cell as { text: unknown }).text;
+  if (!Array.isArray(runs)) {
+    return { usageText: runs as string, metaRaw: null, metaText: null };
+  }
+  const typed = runs as { text: string; color?: string }[];
+  const greyIndexes = typed
+    .map((run, i) => (run.color === META_GREY ? i : -1))
+    .filter((i) => i !== -1);
+  // Production emits at most one grey segment per packed row, and always last in the stream — both
+  // properties are relied on when reading these cells back, so assert them rather than assume.
+  if (greyIndexes.length > 1) {
+    throw new Error(`Expected at most ONE grey meta run per Usage cell, got ${greyIndexes.length}`);
+  }
+  const metaIndex = greyIndexes[0];
+  if (metaIndex === undefined) {
+    return { usageText: typed.map((r) => r.text).join(''), metaRaw: null, metaText: null };
+  }
+  if (metaIndex !== typed.length - 1) {
+    throw new Error(`Grey meta run at index ${metaIndex} of ${typed.length} — expected it last`);
+  }
+  const metaRaw = typed[metaIndex]!.text;
+  return {
+    usageText: typed
+      .slice(0, metaIndex)
+      .map((r) => r.text)
+      .join(''),
+    metaRaw,
+    metaText: metaRaw.replace(/^\n/, ''),
+  };
+}
+
+// Locates the single table row whose Vendor cell reads exactly `vendorName`, and returns its Usage
+// cell split into usage prose vs. trailing grey meta text.
+function usageCellForVendor(
+  body: unknown[][],
+  vendorName: string,
+): { usageText: string; metaText: string | null } {
+  const matches = body.filter(
+    (row) => usageCellText((row[0] as { text?: unknown })?.text) === vendorName,
+  );
+  if (matches.length !== 1) {
+    throw new Error(`Expected exactly 1 row for vendor "${vendorName}", found ${matches.length}`);
+  }
+  const row = matches[0]!;
+  return splitUsageCell(row[row.length - 1]);
+}
+
 // ─── Real fixture: one valid embeddable PDF, fetched for the "split with doc" invoice ─────────
 
 async function makeRealPdfBytes(): Promise<ArrayBuffer> {
@@ -860,32 +929,26 @@ describe('report PDF pipeline — real, unmocked end-to-end render', () => {
           (c) => typeof c === 'object' && c !== null && 'table' in c,
         ) as { table: { body: unknown[][] } };
 
-        // #1929 round 4: attachmentsNote no longer stacks into the usage row's cell — it renders
-        // as its OWN continuation row, immediately after the invoice's usage row (these fixture
-        // invoices carry no budgetLines, so areaText is null and never renders a row in between).
-        // Look up by vendor name rather than a fixed row index, since row counts now vary with
-        // chunking elsewhere in the table.
-        const singleVendorIndex = tableItem.table.body.findIndex(
-          (r) => usageCellText((r[0] as { text?: unknown })?.text) === 'Single Attach Vendor',
-        );
-        const multiVendorIndex = tableItem.table.body.findIndex(
-          (r) => usageCellText((r[0] as { text?: unknown })?.text) === 'Multi Attach Vendor',
-        );
-        expect(singleVendorIndex).toBeGreaterThan(0);
-        expect(multiVendorIndex).toBeGreaterThan(0);
-
-        const singleNoteRow = tableItem.table.body[singleVendorIndex + 1] as { text?: unknown }[];
-        const multiNoteRow = tableItem.table.body[multiVendorIndex + 1] as { text?: unknown }[];
-        expect(usageCellText(singleNoteRow[singleNoteRow.length - 1]!.text)).toBe(expected.single);
-        expect(usageCellText(multiNoteRow[multiNoteRow.length - 1]!.text)).toBe(expected.multi);
+        // #1959: attachmentsNote no longer gets a continuation row of its own — it is the trailing
+        // grey run of the invoice's OWN Usage cell. These fixture invoices carry no budgetLines, so
+        // areaText is null and the meta run holds the attachment note alone (no ' · ' separator).
+        const single = usageCellForVendor(tableItem.table.body, 'Single Attach Vendor');
+        const multi = usageCellForVendor(tableItem.table.body, 'Multi Attach Vendor');
+        expect(single.metaText).toBe(expected.single);
+        expect(multi.metaText).toBe(expected.multi);
+        // The note is genuinely a SEPARATE grey run, not concatenated into the usage prose — so a
+        // regression that inlined it as plain text would fail here rather than pass silently.
+        expect(single.usageText).not.toContain(expected.single);
+        expect(multi.usageText).not.toContain(expected.multi);
       }
     });
 
-    // Story #1923: the "constituted" deposit case no longer produces a footnote at all — it
-    // renders as an inline, unnumbered Deposit label in the allocated cell (a second run, real
-    // i18next `sourceReports.table.attachmentType.deposit` text). Only the "reduced" case still
-    // produces a footnote, now shared/unnumbered (marker `‡`, no vendor/invoice-number prefix).
-    it('renders the real, unnumbered "constituted" Deposit inline label and the real, shared, unnumbered "reduced" footnote in both locales', async () => {
+    // Story #1923 made the "constituted" deposit case an inline, unnumbered Deposit label in the
+    // allocated cell instead of a footnote. #1959 does the same for the remaining two annotations:
+    // the split (`†`) and deposit-reduced (`‡`) FOOTNOTES are gone entirely — buildReportContent
+    // now emits `footnotes: []` and the two cases render as inline grey labels appended to the same
+    // allocated cell, from the real i18next `splitInlineLabel`/`depositReducedInlineLabel` keys.
+    it('#1959: renders the real inline Deposit / split / deposit-reduced labels in the allocated cell — and emits NO † or ‡ footnote anywhere — in both locales', async () => {
       const { buildOverviewContent } = await import('./overviewPdf.js');
       const report = makeUsageFeatureReport();
       const includedIds = new Set(report.invoices.map((inv) => inv.invoiceId));
@@ -896,7 +959,13 @@ describe('report PDF pipeline — real, unmocked end-to-end render', () => {
           formattersFor('en-US'),
           {
             depositLabel: ' (Deposit)',
-            reducedFootnote: '‡: This position reflects deposits claimed separately.',
+            depositReducedLabel: ' (less deposit)',
+            splitLabel: ' (partial)',
+            // The footnote WORDINGS that must no longer appear anywhere in the tree.
+            goneFootnotes: [
+              'This position reflects deposits claimed separately.',
+              'This is a deposit',
+            ],
           },
         ],
         [
@@ -904,8 +973,12 @@ describe('report PDF pipeline — real, unmocked end-to-end render', () => {
           formattersFor('de-DE'),
           {
             depositLabel: ' (Abschlagszahlung)',
-            reducedFootnote:
-              '‡: Diese Position berücksichtigt separat eingereichte Abschlagszahlungen.',
+            depositReducedLabel: ' (abzgl. Abschlag)',
+            splitLabel: ' (Teilbetrag)',
+            goneFootnotes: [
+              'Diese Position berücksichtigt separat eingereichte Abschlagszahlungen.',
+              'Dies ist eine Abschlagszahlung',
+            ],
           },
         ],
       ] as const) {
@@ -922,27 +995,79 @@ describe('report PDF pipeline — real, unmocked end-to-end render', () => {
         const reducedRow = content.rows.find((r) => r.invoiceId === 'inv-deposit-reduced')!;
         expect(reducedRow.isDeposit).toBe(false);
         expect(reducedRow.isDepositReduced).toBe(true);
+        // #1959: the footnote array is empty — the annotations no longer live there at all.
+        expect(content.footnotes).toEqual([]);
 
         const pdfContent = buildOverviewContent(content, new Map(), t);
         const tableItem = pdfContent.find(
           (c) => typeof c === 'object' && c !== null && 'table' in c,
         ) as { table: { body: unknown[][] } };
-        // #1929 round 3: Vendor cell `.text` is now always a run array (buildUsageTextRuns applied
-        // to Vendor, HIGH1) — reconstruct before comparing.
-        const constitutedRowCells = tableItem.table.body.find(
-          (row) => usageCellText((row[0] as { text?: unknown })?.text) === 'Constituted Vendor',
-        ) as { text: unknown | { text: string }[] }[];
-        const allocatedCell = constitutedRowCells[4] as { text: { text: string }[] };
-        expect(Array.isArray(allocatedCell.text)).toBe(true);
-        expect(allocatedCell.text[1]!.text).toBe(expected.depositLabel);
-
-        const notesStack = pdfContent[pdfContent.length - 1] as { stack: { text: string }[] };
-        const texts = notesStack.stack.map((n) => n.text);
-        expect(texts).toContain(expected.reducedFootnote);
-        // No "constituted" wording (footnote form) survives anywhere — it moved to the inline
-        // label above, and the deposit-constituted footnote key/string no longer exists at all.
-        expect(texts.some((text) => text.includes('This is a deposit'))).toBe(false);
-        expect(texts.some((text) => text.includes('Dies ist eine Abschlagszahlung'))).toBe(false);
+
+        // Helper: the allocated cell's runs for the row whose Vendor cell matches. #1929 round 3:
+        // Vendor cell `.text` is always a run array (buildUsageTextRuns applied to Vendor, HIGH1).
+        const allocatedRunsFor = (vendor: string): { text: string }[] => {
+          const cells = tableItem.table.body.find(
+            (row) => usageCellText((row[0] as { text?: unknown })?.text) === vendor,
+          ) as { text: unknown }[];
+          const cell = cells[4] as { text: { text: string }[] };
+          expect(Array.isArray(cell.text)).toBe(true);
+          return cell.text;
+        };
+
+        // Constituted deposit → inline Deposit label as the run right after the amount.
+        expect(allocatedRunsFor('Constituted Vendor')[1]!.text).toBe(expected.depositLabel);
+
+        // #1959: deposit-reduced → inline grey label in the SAME cell (was the `‡` footnote).
+        const reducedRuns = allocatedRunsFor('Reduced Vendor');
+        expect(reducedRuns.map((r) => r.text)).toContain(expected.depositReducedLabel);
+        expect(reducedRuns[0]!.text).not.toContain('‡');
+
+        // #1959: no `†`/`‡` marker and none of the removed footnote wordings survive ANYWHERE in
+        // the rendered content tree (table cells, footnote stack, headers — the whole thing).
+        const allStrings = collectAllStrings(pdfContent);
+        // Positive control first: the inline labels ARE in this same collected set, proving the
+        // collector actually sees the annotation text before we assert the markers' absence.
+        expect(allStrings.some((s) => s.includes(expected.depositLabel.trim()))).toBe(true);
+        expect(allStrings.some((s) => s.includes(expected.depositReducedLabel.trim()))).toBe(true);
+        expect(allStrings.some((s) => s.includes('†'))).toBe(false);
+        expect(allStrings.some((s) => s.includes('‡'))).toBe(false);
+        for (const gone of expected.goneFootnotes) {
+          expect(allStrings.some((s) => s.includes(gone))).toBe(false);
+        }
+      }
+    });
+
+    it('#1959: a split invoice renders the inline split label in its allocated cell instead of a † marker (both locales)', async () => {
+      const { buildOverviewContent } = await import('./overviewPdf.js');
+      const { report, includedIds } = await makeMixedReport();
+
+      for (const [t, formatters, splitLabel] of [
+        [tEn, formattersFor('en-US'), ' (partial)'],
+        [tDe, formattersFor('de-DE'), ' (Teilbetrag)'],
+      ] as const) {
+        const content = buildReportContent(report, includedIds, 'claim', t, formatters, {
+          includeCoverLetter: false,
+          household: null,
+        });
+        const splitRows = content.rows.filter((r) => r.isSplit);
+        expect(splitRows.length).toBeGreaterThan(0); // the fixture really does contain split rows
+
+        const pdfContent = buildOverviewContent(content, new Map(), t);
+        const tableItem = pdfContent.find(
+          (c) => typeof c === 'object' && c !== null && 'table' in c,
+        ) as { table: { body: unknown[][] } };
+
+        for (const splitRow of splitRows) {
+          const cells = tableItem.table.body.find(
+            (row) =>
+              usageCellText((row[0] as { text?: unknown })?.text) === splitRow.vendor &&
+              usageCellText((row[1] as { text?: unknown })?.text) === splitRow.invoiceNumber,
+          ) as { text: unknown }[];
+          const runs = (cells[4] as { text: { text: string }[] }).text;
+          expect(runs.map((r) => r.text)).toContain(splitLabel);
+        }
+
+        expect(collectAllStrings(pdfContent).some((s) => s.includes('†'))).toBe(false);
       }
     });
   });
@@ -1134,7 +1259,7 @@ describe('report PDF pipeline — real, unmocked end-to-end render', () => {
     // Content[] string) — guarding against a future change reflowing the body into per-token
     // inline runs (the #1929 wordBreak technique used elsewhere in reportPdf/), which would
     // silently destroy \n handling if ever applied here.
-    it('AC 1.2/7.2: a multi-paragraph body with an internal blank line renders as one visual line per explicit line/blank-line, in order, at uniform line height', async () => {
+    it('#1959 (supersedes AC 1.2/7.2 single-block shape): a body with a blank line becomes TWO separate paragraph blocks, in order, each keeping its own internal single-newline breaks', async () => {
       const { buildCoverLetterContent } = await import('./coverLetterPdf.js');
       const { report, includedIds } = await makeMixedReport();
       const formatters = formattersFor('en-US');
@@ -1144,39 +1269,88 @@ describe('report PDF pipeline — real, unmocked end-to-end render', () => {
       });
       // Deliberately short lines (well under the ~515pt printable width at 11pt) so word-wrap
       // never adds an extra visual line beyond the explicit \n boundaries — the only thing under
-      // test is whether explicit line/blank-line structure survives, not wrapping behaviour.
-      const body =
-        'First paragraph, line one.\nFirst paragraph, line two.\n\nSecond paragraph, one line.';
+      // test is whether explicit line/paragraph structure survives, not wrapping behaviour.
+      const para1 = 'First paragraph, line one.\nFirst paragraph, line two.';
+      const para2 = 'Second paragraph, one line.';
+      const body = `${para1}\n\n${para2}`;
       const overrides: ReportContentOverrides = { 'coverLetter.body': body };
       const effective = applyOverrides(baseline, overrides);
 
       const pdfContent = buildCoverLetterContent(effective, tEn);
-      const bodyItem = findBodyItem(pdfContent, 'First paragraph, line one.');
-      // Residual gap the architect flagged: line-count + spacing alone don't rule out a rewrite
-      // that alters the TEXT while preserving both (e.g. a per-token reflow that reassembles the
-      // same number of lines from different words). Pin the exact source text too — same pattern
-      // as the sibling AC 1.3 test below.
-      expect(bodyItem['text']).toBe(body);
+
+      // #1959: the double newline is a PARAGRAPH boundary — the body is emitted as two sibling
+      // blocks, never one node whose text still contains '\n\n'.
+      const item1 = findBodyItem(pdfContent, 'First paragraph, line one.');
+      const item2 = findBodyItem(pdfContent, 'Second paragraph, one line.');
+      expect(item1).not.toBe(item2);
+      // Exact text per block pins the split boundary: a wrong split (or a per-token reflow that
+      // preserved line counts but altered the words) fails here.
+      expect(item1['text']).toBe(para1);
+      expect(item2['text']).toBe(para2);
+      // No surviving node still carries the raw '\n\n' — that is the bug this change fixed.
+      expect(collectAllStrings(pdfContent).some((s) => s.includes('\n\n'))).toBe(false);
+      // Document order: paragraph 1 precedes paragraph 2 in the content array.
+      expect(pdfContent.indexOf(item1 as unknown as Content)).toBeLessThan(
+        pdfContent.indexOf(item2 as unknown as Content),
+      );
+      // Inter-paragraph spacing: the first (non-final) paragraph carries the 8pt trailing margin,
+      // the last carries 32pt — this is what makes the paragraph break visible on the page.
+      expect(item1['margin']).toEqual([0, 0, 0, 8]);
+      expect(item2['margin']).toEqual([0, 0, 0, 32]);
 
       await renderCoverLetterPdfContent(pdfContent);
 
-      // 4 explicit segments: "line one.", "line two.", "" (the blank line), "one line." — each
-      // must resolve to exactly one rendered visual line (none of them wraps).
-      expect(linesRenderedFor(bodyItem)).toBe(4);
+      // Real resolved layout: para1's own single newline still produces 2 visual lines; para2
+      // produces 1. The blank line is now spacing between blocks, NOT a rendered empty line — so
+      // the old single-node "4 lines" count is exactly what must no longer happen.
+      expect(linesRenderedFor(item1)).toBe(2);
+      expect(linesRenderedFor(item2)).toBe(1);
       // The real render must not have rewritten `.text` either (pdfmake only ever ADDS derived
       // measurement/position properties to a text node, per the AC 1.3 test's own comment).
-      expect(bodyItem['text']).toBe(body);
+      expect(item1['text']).toBe(para1);
+      expect(item2['text']).toBe(para2);
+
+      // Within para1, its two lines sit exactly one line-height apart...
+      const p1 = item1['positions'] as { top: number }[];
+      const intraGap = p1[1]!.top - p1[0]!.top;
+      expect(intraGap).toBeGreaterThan(0);
+      // ...and the gap ACROSS the paragraph boundary is strictly larger than that, by the 8pt
+      // margin — the assertion that actually proves the paragraph break is visible rather than
+      // merely structural.
+      const p2 = item2['positions'] as { top: number }[];
+      const interGap = p2[0]!.top - p1[1]!.top;
+      expect(interGap).toBeGreaterThan(intraGap);
+      expect(interGap - intraGap).toBeCloseTo(8, 0);
+    });
 
-      const positions = bodyItem['positions'] as { top: number }[];
-      const gaps = positions.slice(1).map((p, i) => p.top - positions[i]!.top);
-      // All 3 gaps (line1->line2, line2->blank, blank->line3) are the SAME line-height increment
-      // — the blank line contributes exactly one ordinary line's worth of vertical space, neither
-      // collapsed to zero (which would make the blank->line3 gap ~0) nor doubled (which would
-      // make either surrounding gap ~2x the others).
-      for (const gap of gaps) {
-        expect(gap).toBeCloseTo(gaps[0]!, 1);
-      }
-      expect(gaps[0]!).toBeGreaterThan(0);
+    it('#1959: a single-paragraph body (no blank line) still renders as exactly ONE block with the 32pt trailing margin', async () => {
+      const { buildCoverLetterContent } = await import('./coverLetterPdf.js');
+      const { report, includedIds } = await makeMixedReport();
+      const formatters = formattersFor('en-US');
+      const baseline = buildReportContent(report, includedIds, 'claim', tEn, formatters, {
+        includeCoverLetter: true,
+        household,
+      });
+      // Two lines, ONE paragraph — a single newline must NOT be treated as a paragraph boundary.
+      const body = 'Only paragraph, line one.\nOnly paragraph, line two.';
+      const effective = applyOverrides(baseline, { 'coverLetter.body': body });
+
+      const pdfContent = buildCoverLetterContent(effective, tEn);
+      const matching = pdfContent.filter(
+        (c) =>
+          typeof c === 'object' &&
+          c !== null &&
+          typeof (c as { text?: unknown }).text === 'string' &&
+          (c as { text: string }).text.includes('Only paragraph'),
+      );
+      expect(matching).toHaveLength(1);
+      const item = matching[0] as unknown as Record<string, unknown>;
+      expect(item['text']).toBe(body);
+      expect(item['margin']).toEqual([0, 0, 0, 32]);
+
+      await renderCoverLetterPdfContent(pdfContent);
+      // The single newline still yields two visual lines inside the one block.
+      expect(linesRenderedFor(item)).toBe(2);
     });
 
     // ─── #1932 AC 1.3: markup-looking characters reach the real render literally, unparsed ─────
@@ -2110,7 +2284,8 @@ describe('report PDF pipeline — real, unmocked end-to-end render', () => {
         statusText: null,
         invoiceAmountText: '€100.00',
         allocatedAmountValueText: '€100.00',
-        allocatedMarkers: '',
+        isSplit: false,
+        isDepositReduced: false,
         isDeposit: false,
         isRefund: false,
         refundNoteText: '',
@@ -2133,6 +2308,8 @@ describe('report PDF pipeline — real, unmocked end-to-end render', () => {
           usage: 'Usage',
           attachmentsNote: 'Attachments',
           deposit: 'Deposit',
+          splitNote: 'partial',
+          depositReducedNote: 'less deposit',
           source: 'Source',
           sourceType: 'Source Type',
           reference: 'Reference',
@@ -2151,104 +2328,147 @@ describe('report PDF pipeline — real, unmocked end-to-end render', () => {
       };
     }
 
-    // Renders `rowOverrides` as the sole content row and returns the resolved table, split into
-    // [usageRows, areaRows, noteRows, summaryRow] by EXPECTED chunk count (derived from the same
-    // splitIntoPageSafeChunks the production code itself uses, not a re-guessed row count) — this
-    // is what lets the assertions below prove reconstruction PER FIELD, not just "some text
-    // somewhere in the tree" (which round 3's own bug would have passed, since the dropped
-    // content's row simply never rendered at all — a shorter body, not corrupted content).
+    // The Usage-cell content stream production builds for a row, mirrored here so the expected row
+    // count can be derived from `packUsageCellRows` — the SAME packer production uses — instead of
+    // from `usageText` alone.
+    //
+    // Why this matters (fix-round regression): the previous version of this helper derived the
+    // expected row count from `splitIntoPageSafeChunks(usageText)`, i.e. it hard-coded the very
+    // assumption the bug lived in — "the grey meta suffix never adds rows". Any correct fix MUST
+    // add rows (16,000 characters of suffix cannot fit one page-safe row), so the helper threw on
+    // its own row-count assertion before the test reached its page-count assertion. The
+    // `it.failing` tripwire below then stayed green for the wrong reason. Deriving from the packer
+    // keeps the helper honest about row COUNT while the per-test assertions below prove the
+    // page-safety bound and losslessness against the INPUT, not against the packer.
+    function cellSegmentsFor(row: ReportContentRow): UsageCellSegment[] {
+      const metaPieces: string[] = [];
+      if (row.areaText) metaPieces.push(row.areaText);
+      if (row.attachmentsNote) metaPieces.push(row.attachmentsNote);
+      const segments: UsageCellSegment[] = [{ text: row.usageText }];
+      if (metaPieces.length > 0) {
+        segments.push({ text: `\n${metaPieces.join(' · ')}`, meta: true });
+      }
+      return segments;
+    }
+
+    // Renders `rowOverrides` as the sole content row through the real, unmocked pdfmake pipeline
+    // and returns everything needed to prove content survival PER CHANNEL.
     async function renderCellScopeRow(rowOverrides: Partial<ReportContentRow>): Promise<{
-      usageRows: unknown[][];
-      areaRows: unknown[][];
-      noteRows: unknown[][];
-      totalDataRows: number;
+      /** Every data row of the group: the first (with leading/amount cells) plus continuations. */
+      dataRows: unknown[][];
+      /** Usage prose reconstructed across ALL rows, excluding every grey suffix piece. */
+      usageText: string;
+      /** Grey suffix reconstructed across ALL rows, leading '\n' stripped. Null when absent. */
+      metaText: string | null;
+      /** Indexes (within dataRows) of the rows carrying a grey suffix piece. */
+      metaRowIndexes: number[];
+      /** Per-row rendered character count of the Usage cell — the packed page-safety budget. */
+      perRowCharCounts: number[];
+      /** Real page count of the rendered PDF. */
+      pageCount: number;
     }> {
-      const { buildOverviewContent } = await import('./overviewPdf.js');
+      const { buildOverviewContent, packUsageCellRows } = await import('./overviewPdf.js');
       const content = makeCellScopeContent(rowOverrides);
       const row = content.rows[0]!;
       const pdfContent = buildOverviewContent(content, new Map(), tEn);
-      await renderOverviewPdfContent(
+      const blob = await renderOverviewPdfContent(
         pdfContent,
         { tableTitle: content.tableTitle, sourceName: content.sourceInfo.sourceName },
         tEn,
       );
       const tableItem = findTableItem(pdfContent);
 
-      const usageChunkCount = splitIntoPageSafeChunks(
-        row.usageText,
+      const packedRowCount = packUsageCellRows(
+        cellSegmentsFor(row),
         MAX_SAFE_USAGE_CHUNK_CHARS,
       ).length;
-      const areaChunkCount = row.areaText
-        ? splitIntoPageSafeChunks(row.areaText, MAX_SAFE_SMALL_CHUNK_CHARS).length
-        : 0;
-      const noteChunkCount = row.attachmentsNote
-        ? splitIntoPageSafeChunks(row.attachmentsNote, MAX_SAFE_SMALL_CHUNK_CHARS).length
-        : 0;
-      const totalDataRows = usageChunkCount + areaChunkCount + noteChunkCount;
-
-      // header (1) + totalDataRows + summary (1) — if this length check fails, a row was
+
+      // header (1) + packedRowCount + summary (1) — if this length check fails, a row was
       // SILENTLY DROPPED (round 3's exact failure mode: the row that couldn't fit the page never
       // rendered at all, rather than throwing or visibly truncating).
-      expect(tableItem.table.body).toHaveLength(1 + totalDataRows + 1);
+      expect(tableItem.table.body).toHaveLength(1 + packedRowCount + 1);
 
-      const dataRows = tableItem.table.body.slice(1, 1 + totalDataRows) as unknown[][];
+      const dataRows = tableItem.table.body.slice(1, 1 + packedRowCount) as unknown[][];
+      const parts = dataRows.map((r) => splitUsageCell(r[r.length - 1]));
+      const metaRaw = parts.map((p) => p.metaRaw ?? '').join('');
+      const pdfDoc = await PDFDocument.load(await blob.arrayBuffer());
       return {
-        usageRows: dataRows.slice(0, usageChunkCount),
-        areaRows: dataRows.slice(usageChunkCount, usageChunkCount + areaChunkCount),
-        noteRows: dataRows.slice(usageChunkCount + areaChunkCount),
-        totalDataRows,
+        dataRows,
+        usageText: parts.map((p) => p.usageText).join(''),
+        metaText: parts.some((p) => p.metaRaw !== null) ? metaRaw.replace(/^\n/, '') : null,
+        metaRowIndexes: parts.flatMap((p, i) => (p.metaRaw !== null ? [i] : [])),
+        perRowCharCounts: parts.map((p) => p.usageText.length + (p.metaRaw?.length ?? 0)),
+        pageCount: pdfDoc.getPageCount(),
       };
     }
 
-    function reconstructUsageColumn(rows: unknown[][]): string {
-      return rows
-        .map((r) => usageCellText((r[r.length - 1] as { text?: unknown })?.text ?? ''))
-        .join('');
+    // The invariant that actually makes every emitted row page-safe under `dontBreakRows: true`:
+    // NO row's Usage cell may exceed the one packed character budget. Asserted against the
+    // constant, not against the packer's own output, so a packer regression cannot satisfy it.
+    function expectEveryRowWithinPageSafeBudget(perRowCharCounts: number[]): void {
+      expect(perRowCharCounts.length).toBeGreaterThan(0);
+      for (const count of perRowCharCounts) {
+        expect(count).toBeLessThanOrEqual(MAX_SAFE_USAGE_CHUNK_CHARS);
+      }
     }
 
-    it('[architect-measured case: usageText 700 + attachmentsNote 400 = 665.8pt] every character of BOTH fields is recoverable — the exact combination round 3 silently dropped', async () => {
-      const usageText = proseOfLength(700);
+    it('[architect-measured case: usageText 700 + attachmentsNote 400] every character of BOTH channels is recoverable, every row stays within the page-safe budget, and the suffix lands on the LAST row', async () => {
+      const usage = proseOfLength(700);
       const attachmentsNote = proseOfLength(400);
-      expect(usageText.length).toBe(700);
+      expect(usage.length).toBe(700);
       expect(attachmentsNote.length).toBe(400);
 
-      const { usageRows, areaRows, noteRows } = await renderCellScopeRow({
-        usageText,
-        attachmentsNote,
-      });
-      expect(areaRows).toHaveLength(0);
-      expect(usageRows.length).toBeGreaterThan(1); // 700 > MAX_SAFE_USAGE_CHUNK_CHARS (650)
-      expect(reconstructUsageColumn(usageRows)).toBe(usageText);
-      expect(reconstructUsageColumn(noteRows)).toBe(attachmentsNote);
+      const { dataRows, usageText, metaText, metaRowIndexes, perRowCharCounts } =
+        await renderCellScopeRow({ usageText: usage, attachmentsNote });
+
+      // 700 + 401 = 1101 characters cannot fit one 650-char page-safe row, so the fix MUST add a
+      // row here — the pre-fix code emitted the suffix on row 0 regardless of the combined size.
+      expect(dataRows.length).toBeGreaterThan(1);
+      // I1 losslessness, asserted against the INPUTS (not against the packer).
+      expect(usageText).toBe(usage);
+      expect(metaText).toBe(attachmentsNote); // no areaText, so no ' · ' separator
+      expectEveryRowWithinPageSafeBudget(perRowCharCounts);
+      // #1959 fix-round placement rule: the suffix trails the prose, so its last piece sits on the
+      // LAST row of the group — never mid-prose with more usage text below it.
+      expect(Math.max(...metaRowIndexes)).toBe(dataRows.length - 1);
     });
 
-    it('[architect-measured case: usageText 700 + 20-leaf-area areaText = 691.0pt] every character of BOTH fields is recoverable, including the aggregate-unbounded areaText', async () => {
-      const usageText = proseOfLength(700);
+    it('[architect-measured case: usageText 700 + 20-leaf-area areaText] every character of BOTH channels is recoverable across the several rows the aggregate-unbounded areaText now spans', async () => {
+      const usage = proseOfLength(700);
       const areaText = twentyLeafAreaText();
-      expect(usageText.length).toBe(700);
+      expect(usage.length).toBe(700);
       expect(areaText.length).toBeGreaterThan(600); // genuinely long — 20 leaf areas, not a stub
 
-      const { usageRows, areaRows, noteRows } = await renderCellScopeRow({
-        usageText,
-        areaText,
-      });
-      expect(noteRows).toHaveLength(0);
-      expect(reconstructUsageColumn(usageRows)).toBe(usageText);
-      expect(reconstructUsageColumn(areaRows)).toBe(areaText);
+      const { dataRows, usageText, metaText, metaRowIndexes, perRowCharCounts } =
+        await renderCellScopeRow({ usageText: usage, areaText });
+
+      expect(usageText).toBe(usage);
+      expect(metaText).toBe(areaText);
+      expectEveryRowWithinPageSafeBudget(perRowCharCounts);
+      // This areaText alone is ~5.8x a page-safe row, so the suffix genuinely spans SEVERAL rows —
+      // the case the pre-fix code crammed into one unbreakable row and then lost.
+      expect(metaRowIndexes.length).toBeGreaterThan(1);
+      expect(Math.max(...metaRowIndexes)).toBe(dataRows.length - 1);
     });
 
     it('[architect-measured case: attachmentsNote 2000 alone = 1119.4pt] every character is recoverable, AND the real page count reflects genuine multi-page need — the direct negation of "rows needing 9 pages rendered as 2"', async () => {
       const attachmentsNote = proseOfLength(2000);
       expect(attachmentsNote.length).toBe(2000);
 
-      const { usageRows, areaRows, noteRows } = await renderCellScopeRow({
-        usageText: 'x', // trivial — isolates attachmentsNote as the sole large field
-        attachmentsNote,
-      });
-      expect(areaRows).toHaveLength(0);
-      expect(usageRows).toHaveLength(1);
-      expect(noteRows.length).toBeGreaterThan(1); // 2000 >> MAX_SAFE_SMALL_CHUNK_CHARS (450)
-      expect(reconstructUsageColumn(noteRows)).toBe(attachmentsNote);
+      const { dataRows, usageText, metaText, metaRowIndexes, perRowCharCounts } =
+        await renderCellScopeRow({
+          usageText: 'x', // trivial — isolates attachmentsNote as the sole large field
+          attachmentsNote,
+        });
+      // A trivial usageText contributes nothing, so the row count is driven entirely by the
+      // suffix: 2001 characters over a 650-char budget genuinely needs several rows. The pre-fix
+      // code emitted ONE row here and pdfmake discarded most of it.
+      expect(dataRows.length).toBeGreaterThan(1);
+      expect(usageText).toBe('x');
+      // Every character of the 2000-char note is recoverable across those rows.
+      expect(metaText).toBe(attachmentsNote);
+      expectEveryRowWithinPageSafeBudget(perRowCharCounts);
+      expect(Math.max(...metaRowIndexes)).toBe(dataRows.length - 1);
 
       // Page-count saturation was the architect's own tell for the silent drop (a row needing 9
       // pages rendered as 2 under round 3). 1119.4pt of attachmentsNote ALONE, at the architect's
@@ -2280,6 +2500,92 @@ describe('report PDF pipeline — real, unmocked end-to-end render', () => {
       const pdfDoc = await PDFDocument.load(await result.blob.arrayBuffer());
       expect(pdfDoc.getPageCount()).toBeGreaterThanOrEqual(2);
     });
+
+    // ─── #1959: page-count saturation regression guard (was the OPEN DEFECT tripwire) ───────────
+    //
+    // The three tests below are a matched set, and must be read together. They assert against
+    // RENDERED OUTPUT (real page counts), deliberately, because the sibling tests above inspect the
+    // pdfmake content tree — and the content tree is exactly where the dropped text still LOOKED
+    // present while the reader received a truncated PDF. Tree-level assertions cannot see this
+    // class of bug at all; only page count can.
+    //
+    // History. #1929 round 4 bounded every row's Usage-cell height by giving
+    // `areaText`/`attachmentsNote` continuation rows of their own. #1959 moved them inline into the
+    // usage cell as a single UNCHUNKED grey run, which re-opened round 4's failure mode: since
+    // `dontBreakRows: true` does not paginate an over-tall row (it silently drops whatever does not
+    // fit — #1929 architect review HIGH 4), and neither field has a maxLength anywhere (editor or
+    // server) while `areaText` is aggregate-unbounded across N leaf areas, ordinary user data could
+    // silently lose pages. MEASURED before the fix, real unmocked renders, 6-column shape:
+    //
+    //   usageText (chunked):  2000ch -> 6 rows / 2 pages    attachmentsNote (inline, pre-fix):
+    //                         4000ch -> 9 rows / 3 pages      2000ch -> 3 rows / 3 pages
+    //                         8000ch -> 15 rows / 5 pages     4000ch -> 3 rows / 2 pages  <- FEWER
+    //                        16000ch -> 27 rows / 9 pages     8000ch -> 3 rows / 2 pages     pages
+    //                                                        16000ch -> 3 rows / 2 pages  <- for
+    //                                                                        MORE content
+    //
+    // `packUsageCellRows` now packs the cell's WHOLE content stream against one budget, so the
+    // suffix channel tracks the usage channel: 16000ch -> 25 suffix rows / 9 pages. The guard below
+    // was an `it.failing` tripwire while the defect was open; production is fixed, so it is a
+    // normal test again. If it ever needs `.failing` back, that is a regression, not a maintenance
+    // step — fix production instead.
+
+    // Positive control: proves the page-count measurement itself is sound, by showing that content
+    // which was never affected (usageText) drives a strictly growing page count over the same size
+    // range. Without this, a saturating page count could be blamed on the measurement.
+    it('control: page count grows strictly with content in the usage channel, proving the page-count measurement is sound', async () => {
+      const counts: number[] = [];
+      for (const length of [2000, 8000, 16000]) {
+        const { pageCount, dataRows } = await renderCellScopeRow({
+          usageText: proseOfLength(length),
+        });
+        expect(dataRows.length).toBeGreaterThan(1);
+        counts.push(pageCount);
+      }
+      // Strictly increasing: 2000 -> 8000 -> 16000 characters each need materially more paper.
+      expect(counts[1]!).toBeGreaterThan(counts[0]!);
+      expect(counts[2]!).toBeGreaterThan(counts[1]!);
+      // And the largest case genuinely needs many pages, not just "more than one".
+      expect(counts[2]!).toBeGreaterThanOrEqual(8);
+    }, 180000);
+
+    it('an unbounded inline attachmentsNote must not saturate the page count — 16,000 characters needs the same ~9 pages it needs as chunked usageText (regression guard for the #1959 silent drop)', async () => {
+      const attachmentsNote = proseOfLength(16000);
+      const { pageCount, dataRows, metaText } = await renderCellScopeRow({
+        usageText: 'x',
+        attachmentsNote,
+      });
+      // The suffix is now packed across many rows rather than crammed into one unbreakable row.
+      expect(dataRows.length).toBeGreaterThan(1);
+      // ...and not one character was traded away for that pagination.
+      expect(metaText).toBe(attachmentsNote);
+      // The same 16,000 characters need 9 pages in the usage channel (see the control above).
+      // Anything close to 2 means pdfmake measured the text and then discarded most of it.
+      expect(pageCount).toBeGreaterThanOrEqual(8);
+    }, 180000);
+
+    // The sharpest formulation of the invariant, and the one that needs no calibrated threshold:
+    // the SAME text must cost the SAME paper regardless of which channel carries it. Pre-fix this
+    // failed spectacularly (9 pages vs 2 at 16,000 characters); it is also immune to future layout
+    // tuning, since both sides move together.
+    it.each([2000, 8000, 16000])(
+      'page count is channel-independent: %i characters costs the same number of pages via attachmentsNote as via usageText',
+      async (length) => {
+        const text = proseOfLength(length);
+        const viaUsage = await renderCellScopeRow({ usageText: text });
+        const viaNote = await renderCellScopeRow({ usageText: 'x', attachmentsNote: text });
+
+        // Both channels round-trip their content losslessly first — otherwise "same page count"
+        // could be satisfied by both channels dropping the same amount.
+        expect(viaUsage.usageText).toBe(text);
+        expect(viaNote.metaText).toBe(text);
+        // The paper cost matches. (Row counts differ by at most one: the suffix carries a leading
+        // '\n' and shares row 0 with the placeholder 'x'.)
+        expect(viaNote.pageCount).toBe(viaUsage.pageCount);
+        expect(Math.abs(viaNote.dataRows.length - viaUsage.dataRows.length)).toBeLessThanOrEqual(1);
+      },
+      300000,
+    );
   });
 
   // ─── #1929 round 2: AC13 header-band smoke test (scenario 20) ─────────────────────────────────
diff --git a/e2e/pages/ReportWizardPage.ts b/e2e/pages/ReportWizardPage.ts
index d79575357..1c013003c 100644
--- a/e2e/pages/ReportWizardPage.ts
+++ b/e2e/pages/ReportWizardPage.ts
@@ -78,9 +78,11 @@
  *     wrapper avoids the shared "table" class-substring collision) with one row per invoice:
  *     read-only vendor/invoiceNumber/date/status-badge/amounts cells, plus a dense (unlabelled)
  *     `usageText` `EditableField` (accessible name `Usage text for {{vendor}},
- *     {{invoiceNumber}}` — `sourceReports.editable.usageTextAriaLabel`) and, only when
- *     `row.attachmentsNote !== null`, an `attachmentsNote` `EditableField` (`Attachments note
- *     for {{vendor}}, {{invoiceNumber}}`). Each `EditableField` shows a `[class*="editedDot"]`
+ *     {{invoiceNumber}}` — `sourceReports.editable.usageTextAriaLabel`). As of Issue #1959 the
+ *     `usageText` field is the row's ONLY editable field: the separate `attachmentsNote`
+ *     `EditableField`/column is gone (see the Issue #1959 paragraph below), so a row contains
+ *     exactly one `textbox` — a useful crisp guard against that column creeping back. Each
+ *     `EditableField` shows a `[class*="editedDot"]`
  *     indicator (a DOM SIBLING of the input within `[class*="fieldWrapper"]`, conditionally
  *     MOUNTED not just opacity-toggled — see `hasEditedIndicator`) and a `[class*="resetButton"]`
  *     (a DOM SIBLING of `fieldWrapper` itself, one level further up — see `resetField`) only
@@ -92,13 +94,13 @@
  *     Invoice Number, Date, Status, Invoice Amount, Allocated Amount) no longer use `<label>`
  *     elements — they render as a `[class*="mobileCardCaption"]` span (the caption) followed by
  *     a `[class*="mobileCardValue"]` span (the value); there is nothing to `getByLabel()` for
- *     these. The EDITABLE rows (`usageText`/`attachmentsNote` `EditableField`s), conversely,
- *     now DO get a real, visible `label` prop (unlike the desktop table's dense/unlabelled
+ *     these. The EDITABLE row (`usageText`, the only one left as of Issue #1959), conversely,
+ *     now DOES get a real, visible `label` prop (unlike the desktop table's dense/unlabelled
  *     copy) — see `mobileUsageField()`'s docstring for why that changes its accessible-name
  *     lookup strategy. Both the desktop `<table>` and the mobile card list are ALWAYS in the
  *     DOM simultaneously (CSS `display:none` toggles per viewport, same convention as the
- *     `WizardStepper` desktop/mobile trees above) — `usageField()`/`attachmentsNoteField()`
- *     stay correctly scoped to `contentTable` and therefore never collide with the mobile
+ *     `WizardStepper` desktop/mobile trees above) — `usageField()`
+ *     stays correctly scoped to `contentTable` and therefore never collides with the mobile
  *     copy, but any NEW mobile-specific locator (`mobileCard()`/`mobileUsageField()` below)
  *     must likewise stay scoped to `mobileCardList`, or it will strict-mode-collide with the
  *     desktop table's copy. **Follow-up bug (#1908)**: `.mobileCardList`
@@ -201,23 +203,46 @@
  * - `sourceInfoBlock` (declared above) is now conditionally rendered — `{!content.isClaim && (
  *   ...)}` — and ABSENT from the DOM entirely for `claim` reports (AC3.1), not merely hidden.
  *   `budget-overview`/`proof-of-funds` reports are unaffected (still render it).
- * - `allocatedMarkers` (the `†`/`‡` text appended after the Allocated Amount, still plain text —
- *   no dedicated locator, read via the row/card's own text) is now SHARED/unnumbered per report
- *   — at most one `†` (any split row) and one `‡` (any deposit-reduced row) — never `†1`/`†2`/
- *   per-invoice numbering. `footnotesBlock`/`footnoteItems` (declared above) mirror this: at
- *   most 2 `<li>` entries total, each `{marker}: {text}` with NO `Vendor (Invoice No.) — `
- *   prefix (dropped — the note is no longer invoice-specific).
+ * - The `†` (split) / `‡` (deposit-reduced) markers this story made shared/unnumbered were
+ *   REPLACED WHOLESALE by inline labels in Issue #1959 — see that paragraph below. Neither glyph
+ *   appears anywhere in the UI or the PDF any more, and `footnotesBlock`/`footnoteItems`
+ *   (declared above) consequently have no producer left.
  * - A constituted-deposit row (the allocation is made up entirely by a deposit tagged to the
- *   reported source) carries NO marker at all — instead an inline `Badge` (`depositBadge`/
+ *   reported source) carries NO marker/label at all — instead an inline `Badge` (`depositBadge`/
  *   `mobileDepositBadge` below) reading "Deposit"/"Abschlagszahlung". There is correspondingly
  *   no "This is a deposit." footnote anymore (`depositConstitutedFootnote` key removed).
  * - `summaryTable`/`summaryTableRows` (declared above) now contains exactly ONE row — `Total`/
  *   `Gesamt` — regardless of how many distinct invoice statuses are included; the old
  *   per-status `Outstanding`/`Paid`/`Quotation`/`Claimed Subtotal` rows are gone
  *   (`sourceReports.table.subtotal` key removed).
- * - `usageAreaText`/`mobileUsageAreaText` below: a read-only leaf-area sub-line rendered under
- *   a row's Usage field, only when the linked item(s) resolve to an area — see the method's own
- *   docstring for why it can never be silently dropped by AI-generated usage text.
+ * - The read-only leaf-area text this story put under a row's Usage field still exists, but it
+ *   is now merged into the combined meta line covered by `usageMetaText()`/
+ *   `mobileUsageMetaText()` below (Issue #1959).
+ *
+ * Issue #1959: report PDF/UI polish — inline meta, inline split/deposit labels, column toggles.
+ * - `†`/`‡` markers and the footnote LIST are GONE. `buildReportContent.ts` no longer pushes any
+ *   `ReportContentFootnote` at all (`ReportContentRow.allocatedMarkers` was replaced by
+ *   `isSplit`/`isDepositReduced` booleans), so `footnotesBlock`/`footnoteItems` are retained
+ *   below purely as negative guards — they can never be populated by the current code path, and
+ *   any test asserting a marker glyph or footnote `<li>` is asserting a superseded design.
+ *   Instead, a split row appends a grey inline `<span class*="inlineNote">` reading `(partial)`
+ *   (de: `(Teilbetrag)`, `sourceReports.table.splitInlineLabel`) INSIDE the Allocated Amount
+ *   cell, and a deposit-reduced row one reading `(less deposit)` (de: `(abzgl. Abschlag)`,
+ *   `depositReducedInlineLabel`) — same cell, same pattern as the existing inline Deposit badge.
+ *   See `inlineNote()`/`mobileInlineNote()` below; the desktop copy carries a leading space
+ *   inside the span (` (partial)`), which `toHaveText`'s whitespace normalization absorbs.
+ * - `row.areaText` and `row.attachmentsNote` are no longer a sub-row and a separate table
+ *   column: both now render as ONE grey read-only meta line inside the Usage cell,
+ *   `[areaText, attachmentsNote].filter(Boolean).join(' · ')` — see `usageMetaText()`/
+ *   `mobileUsageMetaText()` below. The `attachmentsNote` `EditableField` is gone entirely (it is
+ *   read-only text now, no `Attachments note for …` textbox, no `Attachments Note` `<th>`), and
+ *   `overrideKey.row(id).attachmentsNote` is no longer reachable from the UI.
+ * - A `role="group"` (`sourceReports.editable.columnVisibilityLabel` = "Show/hide columns")
+ *   of per-column checkboxes sits next to the table heading, one per column label, all checked
+ *   by default; unchecking one drops that `<th>`/`<td>` from BOTH the desktop table and the
+ *   mobile cards. Local `useState` only — nothing persisted, so it resets on every remount.
+ *   Not currently covered by any locator/scenario here (coverage gap, deliberately not added on
+ *   the #1959 critical path since a new test case reshuffles E2E shard membership).
  *
  * Issue #1932: cover letter overhaul — formatted body, editable signature block, personal
  * sender, professional PDF layout.
@@ -441,11 +466,14 @@ export class ReportWizardPage {
 
   // Story #1923: report table cleanup — total-only summary table (`.summaryTable`, a sibling
   // `<table>` OUTSIDE `.tableWrapper` — see `contentTable`'s own docstring for why that scoping
-  // avoids a substring collision) and the shared footnotes block (`.footnotes`, now at most 2
-  // `<li>` entries — one per marker (`†`/`‡`) — never one per invoice; see `depositBadge`'s
-  // docstring below for the constituted-deposit case, which carries NO marker at all).
+  // avoids a substring collision).
   readonly summaryTable: Locator;
   readonly summaryTableRows: Locator;
+  // The footnotes block (`.footnotes` / its `<li>` entries). As of Issue #1959 NOTHING populates
+  // `content.footnotes` any more (the `†`/`‡` split + deposit-reduced entries became inline
+  // labels — see `inlineNote()`), so the block is never rendered. Kept purely as a NEGATIVE
+  // guard: a scenario asserting `toHaveCount(0)` / `not.toBeVisible()` here is asserting that
+  // the superseded footnote mechanism has not come back. Never assert a positive count on these.
   readonly footnotesBlock: Locator;
   readonly footnoteItems: Locator;
 
@@ -1067,17 +1095,6 @@ export class ReportWizardPage {
     });
   }
 
-  /**
-   * The `attachmentsNote` `EditableField` for a table row — only present when
-   * `row.attachmentsNote !== null` (i.e. the invoice has linked documents). Same
-   * substring-name-matching note as `usageField`.
-   */
-  attachmentsNoteField(vendorName: string, invoiceNumber: string): Locator {
-    return this.contentTableRow(vendorName, invoiceNumber).getByRole('textbox', {
-      name: `Attachments note for ${vendorName}, ${invoiceNumber}`,
-    });
-  }
-
   /**
    * A single mobile-card row within `mobileCardList` (fixed #1904) — a DIRECT CHILD
    * (`>` combinator) so this never also matches a nested `[class*="mobileCardRow"]`
@@ -1117,8 +1134,10 @@ export class ReportWizardPage {
    * The inline "Deposit" `Badge` (`[class*="depositBadge"]`, composed from the shared
    * `.attachmentDeposit` variant) rendered in a desktop content-table row's Allocated Amount
    * cell when `row.isDeposit` — a constituted-deposit row (AC2.1), i.e. the row's allocation is
-   * made up entirely by a deposit tagged to the CURRENTLY reported source. Carries NO `†`/`‡`
-   * marker of its own. Scoped to the row so it never collides with `mobileDepositBadge`'s copy
+   * made up entirely by a deposit tagged to the CURRENTLY reported source. Carries no inline
+   * note of its own (see `inlineNote()` — neither `(partial)` nor `(less deposit)`, and, before
+   * Issue #1959 replaced them, neither `†` nor `‡`).
+   * Scoped to the row so it never collides with `mobileDepositBadge`'s copy
    * (both trees share the `depositBadge` class and are always in the DOM simultaneously, per
    * the class docstring's dual-DOM-tree convention).
    */
@@ -1132,20 +1151,47 @@ export class ReportWizardPage {
   }
 
   /**
-   * The read-only, non-editable leaf-area sub-line (`[class*="usageAreaText"]`) rendered below
-   * a desktop row's `usageText` field when `row.areaText` is non-null (AC5.2) — distinct
-   * comma-joined leaf area names, never concatenated into the editable `usageText` string
-   * itself (AC5.3), so it survives both manual edits and AI-generated usage text overwriting
-   * `usageText` (`applyAiContent.ts` only ever assigns `row.usageText`, never `row.areaText`).
-   * Absent entirely (not an empty element) when the row has no area (AC5.4).
+   * The read-only, non-editable grey meta line (`[class*="usageMetaText"]`) rendered inside a
+   * desktop row's Usage CELL, directly below its `usageText` field. As of Issue #1959 this is
+   * the single home for BOTH `row.areaText` (comma-joined leaf area names, AC5.2) and
+   * `row.attachmentsNote` (e.g. "1 attachment: Invoice") — rendered as
+   * `[areaText, attachmentsNote].filter(Boolean).join(' · ')`, so a row with both shows
+   * `Kitchen · 1 attachment: Invoice`. It replaces BOTH the old `.usageAreaText` sub-line and
+   * the old separate, editable `Attachments Note` column (removed — see the Issue #1959
+   * paragraph in the class docstring).
+   *
+   * Neither half is ever concatenated into the editable `usageText` string itself (AC5.3), so
+   * the meta line survives manual edits AND AI-generated usage text (`applyAiContent.ts` only
+   * ever assigns `row.usageText`, never `row.areaText`/`row.attachmentsNote`). Absent entirely
+   * (not an empty element) when the row has neither an area nor a linked document (AC5.4).
+   */
+  usageMetaText(vendorName: string, invoiceNumber: string): Locator {
+    return this.contentTableRow(vendorName, invoiceNumber).locator('[class*="usageMetaText"]');
+  }
+
+  /** The same meta line within a mobile card's Usage row (AC6.2) — a `<span>`, not a `<div>`. */
+  mobileUsageMetaText(vendorName: string, invoiceNumber: string): Locator {
+    return this.mobileCard(vendorName, invoiceNumber).locator('[class*="usageMetaText"]');
+  }
+
+  /**
+   * The grey inline note(s) (`[class*="inlineNote"]`) appended INSIDE a desktop row's Allocated
+   * Amount cell by Issue #1959, which replaced the `†`/`‡` footnote markers: `(partial)` when
+   * `row.isSplit` (the amount shown is only this source's portion) and `(less deposit)` when
+   * `row.isDepositReduced` (separately-claimed deposits already netted off). A constituted-
+   * deposit row (`row.isDeposit`) gets NEITHER — it gets the inline `depositBadge` instead — so
+   * asserting `toHaveCount(0)` here is a meaningful negative for that case.
+   *
+   * Scoped to the row, so it never collides with `mobileInlineNote`'s copy (both DOM trees are
+   * always present simultaneously — see the class docstring's dual-DOM-tree convention).
    */
-  usageAreaText(vendorName: string, invoiceNumber: string): Locator {
-    return this.contentTableRow(vendorName, invoiceNumber).locator('[class*="usageAreaText"]');
+  inlineNote(vendorName: string, invoiceNumber: string): Locator {
+    return this.contentTableRow(vendorName, invoiceNumber).locator('[class*="inlineNote"]');
   }
 
-  /** The same area sub-line within a mobile card's Usage field (AC6.2). */
-  mobileUsageAreaText(vendorName: string, invoiceNumber: string): Locator {
-    return this.mobileCard(vendorName, invoiceNumber).locator('[class*="usageAreaText"]');
+  /** The same inline note(s) within a mobile card's Allocated Amount value. */
+  mobileInlineNote(vendorName: string, invoiceNumber: string): Locator {
+    return this.mobileCard(vendorName, invoiceNumber).locator('[class*="inlineNote"]');
   }
 
   /** Fills an `EditableField` (input or textarea) with `value`, firing its `onChange`. */
diff --git a/e2e/tests/budget/reportWizardAiGeneration.spec.ts b/e2e/tests/budget/reportWizardAiGeneration.spec.ts
index 18babf29b..5c8fc8a76 100644
--- a/e2e/tests/budget/reportWizardAiGeneration.spec.ts
+++ b/e2e/tests/budget/reportWizardAiGeneration.spec.ts
@@ -48,10 +48,12 @@
  * - Scenario 7: A confirmed Step 1-4 change (via the existing discard-confirm modal) clears
  *   previously-generated AI content, same as it clears manual overrides — the fields revert to
  *   the plain derived (#1898/#1900) baseline and the provenance note disappears.
- * - Scenario 8 (Story #1923 AC5.3): the read-only area sub-line under a row's Usage field
- *   survives AI generation — `applyAiContent.ts` only ever assigns `row.usageText`, never
- *   `row.areaText`, so a row whose linked item has an assigned area keeps showing that area
- *   after "Enhance with AI" overwrites the usage text itself.
+ * - Scenario 8 (Story #1923 AC5.3): the read-only area meta text in a row's Usage cell survives
+ *   AI generation — `applyAiContent.ts` only ever assigns `row.usageText`, never `row.areaText`,
+ *   so a row whose linked item has an assigned area keeps showing that area after "Enhance with
+ *   AI" overwrites the usage text itself. Issue #1959 moved that text from its own
+ *   `.usageAreaText` sub-line into the combined `.usageMetaText` line (area + attachments note)
+ *   inside the same cell; the survival property it asserts is unchanged.
  * - Scenario 9 (Issue #1931 a11y addition): the button carries `aria-describedby` pointing at a
  *   visually-hidden sibling span with the expected overwrite-behavior text, rendered
  *   unconditionally alongside the button (present with no manual edits yet, not only once a
@@ -747,11 +749,11 @@ test.describe('Report wizard AI generation — discard clears AI content (Scenar
 });
 
 // ─────────────────────────────────────────────────────────────────────────────
-// Scenario 8: The area sub-line survives AI generation (Story #1923 AC5.3)
+// Scenario 8: The area meta text survives AI generation (Story #1923 AC5.3)
 // ─────────────────────────────────────────────────────────────────────────────
 
-test.describe('Report wizard AI generation — area sub-line survives generation (Scenario 8)', () => {
-  test('A row\'s read-only area sub-line is still present after "Enhance with AI" overwrites the usage text, because applyAiContent only ever assigns usageText, never areaText', async ({
+test.describe('Report wizard AI generation — area meta text survives generation (Scenario 8)', () => {
+  test('A row\'s read-only area meta text is still present after "Enhance with AI" overwrites the usage text, because applyAiContent only ever assigns usageText, never areaText', async ({
     page,
     testPrefix,
   }) => {
@@ -794,11 +796,15 @@ test.describe('Report wizard AI generation — area sub-line survives generation
       const vendorName = `${testPrefix} AiArea Vendor`;
       await reachStep5WithAiConfigured(wizard, sourceId);
       const usage = wizard.usageField(vendorName, invoice.invoiceNumber!);
-      const areaLine = wizard.usageAreaText(vendorName, invoice.invoiceNumber!);
+      // Issue #1959: the area now renders in the row's combined read-only meta line
+      // (`.usageMetaText`, area + attachments note) inside the Usage cell rather than its own
+      // `.usageAreaText` sub-row. This invoice has no linked document, so the meta line is the
+      // area name alone.
+      const areaLine = wizard.usageMetaText(vendorName, invoice.invoiceNumber!);
       const areaName = `${testPrefix} Bathroom`;
 
       // Before generation: the derived baseline usage text (linked item name) alongside the
-      // area sub-line.
+      // area meta text.
       await expect(areaLine).toBeVisible();
       await expect(areaLine).toHaveText(areaName);
 
@@ -806,11 +812,13 @@ test.describe('Report wizard AI generation — area sub-line survives generation
       await expect(usage).toHaveValue('AI-generated usage description overwriting the row');
       await expect(wizard.aiGeneratedNote).toBeVisible();
 
-      // After generation: usageText was overwritten by the AI description, but the area
-      // sub-line — a separate, non-editable field never touched by `applyAiContent` — is
-      // unchanged (AC5.3).
+      // After generation: usageText was overwritten by the AI description, but the area meta
+      // text — read-only, never touched by `applyAiContent` — is unchanged (AC5.3). The AI text
+      // itself does not mention the area, so the assertion above cannot be satisfied by the
+      // overwritten usage value bleeding into the meta line.
       await expect(areaLine).toBeVisible();
       await expect(areaLine).toHaveText(areaName);
+      expect(await usage.inputValue()).not.toContain(areaName);
       expect(counter.count).toBe(1);
     } finally {
       if (workItemId) await deleteWorkItemViaApi(page, workItemId);
diff --git a/e2e/tests/budget/reportWizardEditableContent.spec.ts b/e2e/tests/budget/reportWizardEditableContent.spec.ts
index a8c1469c5..1fb615534 100644
--- a/e2e/tests/budget/reportWizardEditableContent.spec.ts
+++ b/e2e/tests/budget/reportWizardEditableContent.spec.ts
@@ -19,7 +19,11 @@
  *   for #1904) must stay hidden at desktop width. It previously lacked a base `display: none`
  *   and duplicated the table; the fix landed in `ReportContentEditor.module.css`.
  * - Scenario 2: Editing a field shows its edited-dot indicator on THAT field only — not on
- *   sibling fields, not on the same field for a different invoice row.
+ *   sibling fields, not on the same field for a different invoice row. Since Issue #1959 the
+ *   table row's only editable field is `usageText` (the `attachmentsNote` `EditableField`/column
+ *   is gone — the note is read-only meta text now), so the sibling-field axis is exercised
+ *   against the cover letter's `subject`/`body`, and the removed field's absence is asserted
+ *   here as "exactly one textbox per row".
  * - Scenario 3: The per-field reset button reverts the value to its generated baseline and
  *   clears the edited-dot.
  * - Scenario 4: A guarded step 1-4 mutation while dirty shows the discard-confirm modal; "Keep
@@ -64,24 +68,33 @@
  *   attach-documents/cover-letter checkboxes on the same step (previously it bypassed the
  *   guard and applied silently).
  *
- * Story #1923 (report table cleanup): shared footnotes, inline deposit label, claim metadata
- * suppression, total-only summary, area in the Usage cell. See `ReportWizardPage.ts`'s class
- * docstring for the full locator reference (`sourceInfoBlock`, `depositBadge`/
- * `mobileDepositBadge`, `usageAreaText`/`mobileUsageAreaText`, `summaryTable`/
- * `summaryTableRows`, `footnotesBlock`/`footnoteItems`).
+ * Story #1923 (report table cleanup): inline deposit label, claim metadata suppression,
+ * total-only summary, area in the Usage cell. Issue #1959 then REPLACED this story's `†`/`‡`
+ * markers + footnote list with inline `(partial)`/`(less deposit)` labels in the Allocated Amount
+ * cell, and merged the area sub-line and the (previously editable, separately-columned)
+ * attachments note into ONE read-only grey meta line in the Usage cell. See `ReportWizardPage.ts`'s
+ * class docstring for the full locator reference (`sourceInfoBlock`, `depositBadge`/
+ * `mobileDepositBadge`, `inlineNote`/`mobileInlineNote`, `usageMetaText`/`mobileUsageMetaText`,
+ * `summaryTable`/`summaryTableRows`, and `footnotesBlock`/`footnoteItems` — retained as
+ * negative-only guards now that nothing populates `content.footnotes`).
  * - Scenario 16: A `claim` report omits the source-info metadata block entirely (AC3.1) — the
  *   counterpart to Scenario 1's `budget-overview` regression guard (AC3.3).
  * - Scenario 17: A constituted-deposit row (the row's allocation is made up entirely by a
  *   deposit tagged to the currently reported source) shows the inline "Deposit" badge on
- *   desktop, tablet, AND mobile, carries no `‡` marker, and the footnotes list has no entry at
- *   all for it (AC2.1, AC2.2).
- * - Scenario 18: Two or more split invoices share exactly ONE unnumbered `†` marker and exactly
- *   one footnote entry, with no vendor/invoice-number prefix (AC1.1-AC1.2).
+ *   desktop, tablet, AND mobile, carries NO inline `(partial)`/`(less deposit)` note (nor either
+ *   legacy `†`/`‡` glyph), and there is no footnotes block at all (AC2.1, AC2.2).
+ * - Scenario 18: Every split invoice carries its OWN inline `(partial)` label in its Allocated
+ *   Amount cell (Issue #1959, superseding AC1.1-AC1.2's shared unnumbered `†` marker), and the
+ *   footnote list — including the long-form "Amount shown reflects only the portion allocated to
+ *   this source." sentence — is gone from the page entirely.
  * - Scenario 19: Invoices spanning two or more statuses still produce exactly one summary row
  *   (`Total`) — no per-status subtotal rows (AC4.1-AC4.2).
  * - Scenario 20: A budget line linked to an item with an assigned area shows the item's leaf
- *   area name as a distinct, read-only line below the Usage field on desktop, tablet, AND
- *   mobile; an item with no area renders no area line and no empty gap (AC5.2, AC5.4, AC5.5).
+ *   area name as read-only grey meta text inside the Usage cell on desktop, tablet, AND mobile —
+ *   joined by " · " with the attachments note when the invoice also has a linked document (Issue
+ *   #1959), never as a separate editable Attachments Note column, and never folded into the
+ *   editable usage text (AC5.2, AC5.3); an invoice with neither renders no meta text and no
+ *   empty gap (AC5.4, AC5.5).
  *
  * Issue #1932 (cover letter overhaul): formatted body, editable signature block, personal
  * sender, professional PDF layout. See `ReportWizardPage.ts`'s own "Issue #1932" docstring
@@ -521,31 +534,44 @@ test.describe('Report wizard editable content — edited-dot placement (Scenario
 
       const vendorName = `${testPrefix} Dot Vendor`;
       const usageA = wizard.usageField(vendorName, invoiceA.invoiceNumber!);
-      const attachmentsNoteA = wizard.attachmentsNoteField(vendorName, invoiceA.invoiceNumber!);
       const usageB = wizard.usageField(vendorName, invoiceB.invoiceNumber!);
       const subject = wizard.letterField('subject');
+      const body = wizard.letterField('body');
+
+      // Invoice A's linked document (seeded above) used to give its row a SECOND editable field
+      // (the `Attachments note for …` `EditableField`) — Issue #1959 removed that column and
+      // turned the note into read-only grey meta text inside the Usage cell. Asserted as a
+      // positive/negative pair so this can't pass vacuously against a row that simply has no
+      // linked document: the note text IS rendered, and the row has exactly ONE textbox (Usage).
+      await expect(wizard.usageMetaText(vendorName, invoiceA.invoiceNumber!)).toHaveText(
+        '1 attachment: Invoice',
+      );
+      await expect(
+        wizard.contentTableRow(vendorName, invoiceA.invoiceNumber!).getByRole('textbox'),
+      ).toHaveCount(1);
 
       // Baseline: nothing edited anywhere.
       expect(await wizard.hasEditedIndicator(usageA)).toBe(false);
-      expect(await wizard.hasEditedIndicator(attachmentsNoteA)).toBe(false);
       expect(await wizard.hasEditedIndicator(usageB)).toBe(false);
       expect(await wizard.hasEditedIndicator(subject)).toBe(false);
+      expect(await wizard.hasEditedIndicator(body)).toBe(false);
 
       // Edit invoice A's usage field only.
       await wizard.editField(usageA, 'Custom usage note A');
       expect(await wizard.hasEditedIndicator(usageA)).toBe(true);
-      // Sibling field on the SAME row: unaffected.
-      expect(await wizard.hasEditedIndicator(attachmentsNoteA)).toBe(false);
       // Same field TYPE on a DIFFERENT row: unaffected.
       expect(await wizard.hasEditedIndicator(usageB)).toBe(false);
-      // A cover-letter field: unaffected.
+      // Cover-letter fields: unaffected.
       expect(await wizard.hasEditedIndicator(subject)).toBe(false);
+      expect(await wizard.hasEditedIndicator(body)).toBe(false);
 
-      // Editing the second field on the same row shows its own dot too, without disturbing
-      // the first.
-      await wizard.editField(attachmentsNoteA, 'Custom attachment note A');
+      // Editing a SIBLING field (now the cover letter's subject — the table row has only one
+      // editable field left) shows its own dot too, without disturbing the first, and still
+      // without leaking onto its own sibling (`body`) or the other row.
+      await wizard.editField(subject, 'Custom subject line');
       expect(await wizard.hasEditedIndicator(usageA)).toBe(true);
-      expect(await wizard.hasEditedIndicator(attachmentsNoteA)).toBe(true);
+      expect(await wizard.hasEditedIndicator(subject)).toBe(true);
+      expect(await wizard.hasEditedIndicator(body)).toBe(false);
       expect(await wizard.hasEditedIndicator(usageB)).toBe(false);
     } finally {
       if (workItemId) await deleteWorkItemViaApi(page, workItemId);
@@ -1463,7 +1489,7 @@ test.describe('Report wizard editable content — report-language guard (Scenari
 // ─────────────────────────────────────────────────────────────────────────────
 
 test.describe('Report wizard editable content — claim reports omit the metadata block (Scenario 16)', () => {
-  test('A claim report hides the source-info metadata block entirely, while the title, table, summary, and (when present) footnotes still render (Story #1923 AC3.1, AC3.4)', async ({
+  test('A claim report hides the source-info metadata block entirely, while the title, table, and summary still render (Story #1923 AC3.1, AC3.4)', async ({
     page,
     testPrefix,
   }) => {
@@ -1510,15 +1536,15 @@ test.describe('Report wizard editable content — claim reports omit the metadat
 });
 
 // ─────────────────────────────────────────────────────────────────────────────
-// Scenario 17: Constituted-deposit row shows the inline Deposit badge, no ‡ marker, no footnote
-// (Story #1923 AC2.1, AC2.2)
+// Scenario 17: Constituted-deposit row shows the inline Deposit badge and no deposit-reduced
+// note (Story #1923 AC2.1, AC2.2; note shape updated by Issue #1959)
 // ─────────────────────────────────────────────────────────────────────────────
 
 test.describe(
   'Report wizard editable content — inline deposit badge (Scenario 17)',
   { tag: '@responsive' },
   () => {
-    test('A row whose allocation is made up entirely by a deposit tagged to the reported source shows an inline "Deposit" badge instead of a ‡ marker, with no matching footnote entry, on desktop, tablet, and mobile', async ({
+    test('A row whose allocation is made up entirely by a deposit tagged to the reported source shows an inline "Deposit" badge and no "(less deposit)" note, with no footnote entry, on desktop, tablet, and mobile', async ({
       page,
       testPrefix,
     }) => {
@@ -1559,8 +1585,10 @@ test.describe(
 
         // View source B's own claim report — this invoice has zero budget lines for B, so its
         // entire row here is the tagged deposit: isSplit (spans A + B) with an empty
-        // `budgetLines` slice for B → constituted-deposit, no † (no budget-line split for B)
-        // and no ‡ (the deposit IS tagged to B, not "reduced").
+        // `budgetLines` slice for B → constituted-deposit, so `isSplit` is false (no
+        // budget-line split for B) and `isDepositReduced` is false (the deposit IS tagged to B,
+        // not "reduced") → NEITHER inline note (Issue #1959 replaced the old †/‡ markers with
+        // the `(partial)`/`(less deposit)` labels asserted against here).
         await reachStep5(wizard, sourceBId);
 
         const vendorName = `${testPrefix} Deposit Vendor`;
@@ -1573,7 +1601,11 @@ test.describe(
           const badge = wizard.mobileDepositBadge(vendorName, invoice.invoiceNumber!);
           await expect(badge).toBeVisible();
           await expect(badge).toHaveText('Deposit');
+          // The badge is the row's ONLY allocated-amount annotation: no `(partial)`/`(less
+          // deposit)` inline note (Issue #1959), and — belt and braces — neither legacy glyph.
+          await expect(wizard.mobileInlineNote(vendorName, invoice.invoiceNumber!)).toHaveCount(0);
           const cardText = (await card.textContent()) ?? '';
+          expect(cardText).not.toContain('†');
           expect(cardText).not.toContain('‡');
         } else {
           await expect(wizard.contentTable).toBeVisible();
@@ -1582,15 +1614,18 @@ test.describe(
           const badge = wizard.depositBadge(vendorName, invoice.invoiceNumber!);
           await expect(badge).toBeVisible();
           await expect(badge).toHaveText('Deposit');
+          await expect(wizard.inlineNote(vendorName, invoice.invoiceNumber!)).toHaveCount(0);
           const rowText = (await row.textContent()) ?? '';
+          expect(rowText).not.toContain('†');
           expect(rowText).not.toContain('‡');
         }
 
         // No footnote entry at all — no split (B has zero budget lines here), no
-        // deposit-reduced (the deposit IS tagged to B, not "reduced"), and the removed
+        // deposit-reduced (the deposit IS tagged to B, not "reduced"), the removed
         // `depositConstitutedFootnote` key ("This is a deposit.") never had a footnote to begin
-        // with even before this story.
-        await expect(wizard.footnotesBlock).not.toBeVisible();
+        // with even before Story #1923, and as of Issue #1959 nothing populates
+        // `content.footnotes` at all any more.
+        await expect(wizard.footnotesBlock).toHaveCount(0);
       } finally {
         if (workItemId) await deleteWorkItemViaApi(page, workItemId);
         if (sourceBId) await deleteBudgetSourceViaApi(page, sourceBId);
@@ -1602,11 +1637,12 @@ test.describe(
 );
 
 // ─────────────────────────────────────────────────────────────────────────────
-// Scenario 18: Two or more split invoices share ONE unnumbered † footnote (Story #1923 AC1)
+// Scenario 18: Every split invoice carries its own inline "(partial)" label and there is no
+// footnote list at all (Story #1923 AC1, superseded by Issue #1959)
 // ─────────────────────────────────────────────────────────────────────────────
 
-test.describe('Report wizard editable content — shared split footnote (Scenario 18)', () => {
-  test('Two split invoices both carry an unnumbered † marker, and the footnote list has exactly one † entry with no vendor/invoice-number prefix', async ({
+test.describe('Report wizard editable content — inline split label (Scenario 18)', () => {
+  test('Two split invoices each show a grey inline "(partial)" note in their Allocated Amount cell, with no †/‡ marker and no footnote list anywhere on the page', async ({
     page,
     testPrefix,
   }) => {
@@ -1658,25 +1694,41 @@ test.describe('Report wizard editable content — shared split footnote (Scenari
       const vendorName = `${testPrefix} Split Vendor`;
       const row1 = wizard.contentTableRow(vendorName, invoice1.invoiceNumber!);
       const row2 = wizard.contentTableRow(vendorName, invoice2.invoiceNumber!);
-      await expect(row1).toContainText('†');
-      await expect(row2).toContainText('†');
 
-      // Unnumbered — no digit ever directly follows the marker glyph.
+      // Issue #1959 replaced the shared, unnumbered `†` marker + its footnote entry with a grey
+      // inline label appended to each split row's Allocated Amount cell. Positive first: exactly
+      // one such note per split row, reading `(partial)` — the row is genuinely split (its own
+      // €100.00 / €150.00 portion of a €300.00 / €400.00 invoice), so this cannot pass on a
+      // mis-seeded, non-split page.
+      const note1 = wizard.inlineNote(vendorName, invoice1.invoiceNumber!);
+      const note2 = wizard.inlineNote(vendorName, invoice2.invoiceNumber!);
+      await expect(note1).toHaveCount(1);
+      await expect(note2).toHaveCount(1);
+      await expect(note1).toHaveText('(partial)');
+      await expect(note2).toHaveText('(partial)');
+      // The label lives in the amount cell, not appended to the editable usage text.
+      await expect(row1).toContainText('€100.00 (partial)');
+      await expect(row2).toContainText('€150.00 (partial)');
+
+      // Negatives, each paired with the positives above: neither footnote glyph survives
+      // anywhere in the row (numbered or not), and the footnote list has no producer left — the
+      // block is absent from the DOM entirely rather than merely empty.
       const row1Text = (await row1.textContent()) ?? '';
       const row2Text = (await row2.textContent()) ?? '';
-      expect(row1Text).not.toMatch(/†\d/);
-      expect(row2Text).not.toMatch(/†\d/);
-
-      // Exactly one shared footnote entry, no per-invoice prefix.
-      await expect(wizard.footnoteItems).toHaveCount(1);
-      const footnoteText = (await wizard.footnoteItems.first().textContent()) ?? '';
-      expect(footnoteText).toMatch(/^†:/);
-      expect(footnoteText).not.toContain(vendorName);
-      expect(footnoteText).not.toContain(invoice1.invoiceNumber!);
-      expect(footnoteText).not.toContain(invoice2.invoiceNumber!);
-      expect(footnoteText).toContain(
+      expect(row1Text).not.toContain('†');
+      expect(row2Text).not.toContain('†');
+      expect(row1Text).not.toContain('‡');
+      expect(row2Text).not.toContain('‡');
+      await expect(wizard.footnotesBlock).toHaveCount(0);
+      await expect(wizard.footnoteItems).toHaveCount(0);
+
+      // The long-form footnote sentence the marker used to point at is gone from the page too
+      // (it now only exists as the short inline label asserted above).
+      const pageText = (await page.locator('main').textContent()) ?? '';
+      expect(pageText).not.toContain(
         'Amount shown reflects only the portion allocated to this source.',
       );
+      expect(pageText).toContain('(partial)');
     } finally {
       if (reportedWorkItemId) await deleteWorkItemViaApi(page, reportedWorkItemId);
       if (otherWorkItemId) await deleteWorkItemViaApi(page, otherWorkItemId);
@@ -1745,14 +1797,15 @@ test.describe('Report wizard editable content — total-only summary (Scenario 1
 });
 
 // ─────────────────────────────────────────────────────────────────────────────
-// Scenario 20: Area sub-line in the Usage cell (Story #1923 AC5)
+// Scenario 20: Area + attachments meta line in the Usage cell (Story #1923 AC5, reshaped by
+// Issue #1959 — one combined grey read-only meta line, no separate Attachments Note column)
 // ─────────────────────────────────────────────────────────────────────────────
 
 test.describe(
-  'Report wizard editable content — area sub-line in Usage cell (Scenario 20)',
+  'Report wizard editable content — area/attachments meta line in Usage cell (Scenario 20)',
   { tag: '@responsive' },
   () => {
-    test('A budget line linked to an item with an assigned area renders the leaf area name as a distinct, read-only line below Usage text (desktop, tablet, mobile); an item with no area renders no area line at all', async ({
+    test('A budget line linked to an item with an assigned area renders the leaf area name as read-only grey meta text inside the Usage cell, joined with the attachments note by " · " when the invoice also has a linked document, and never as a separate editable Attachments Note column (desktop, tablet, mobile); an item with neither renders no meta text at all', async ({
       page,
       testPrefix,
     }) => {
@@ -1800,6 +1853,22 @@ test.describe(
             status: 'pending',
           },
         );
+        // Third row: same area-bearing work item, but this invoice ALSO has a linked document,
+        // so its meta line must carry BOTH halves joined by " · " (Issue #1959 folded the old
+        // separate, editable Attachments Note column into this same line).
+        const invoiceWithBoth = await seedAllocatedInvoice(
+          page,
+          workItemWithAreaId,
+          vendorId,
+          sourceId,
+          {
+            invoiceNumber: `${testPrefix}-AREA-003`,
+            amount: 140,
+            date: '2026-06-22',
+            status: 'pending',
+          },
+        );
+        await linkDocumentToInvoiceViaApi(page, invoiceWithBoth.id, 555020);
 
         await reachStep5(wizard, sourceId);
 
@@ -1807,20 +1876,58 @@ test.describe(
         const areaName = `${testPrefix} Kitchen`;
         const isMobile = test.info().project.name === 'mobile';
 
-        if (isMobile) {
-          const areaLine = wizard.mobileUsageAreaText(vendorName, invoiceWithArea.invoiceNumber!);
-          await expect(areaLine).toBeVisible();
-          await expect(areaLine).toHaveText(areaName);
-          await expect(
-            wizard.mobileUsageAreaText(vendorName, invoiceNoArea.invoiceNumber!),
-          ).toHaveCount(0);
-        } else {
-          const areaLine = wizard.usageAreaText(vendorName, invoiceWithArea.invoiceNumber!);
-          await expect(areaLine).toBeVisible();
-          await expect(areaLine).toHaveText(areaName);
-          await expect(wizard.usageAreaText(vendorName, invoiceNoArea.invoiceNumber!)).toHaveCount(
-            0,
-          );
+        // Only the container/locator flavour differs per viewport — the desktop meta line is a
+        // `<div>` in the Usage `<td>`, the mobile one a `<span>` in the Usage card row; both are
+        // `[class*="usageMetaText"]` and both are read-only, so the assertions themselves are
+        // identical for all three projects.
+        const metaOf = (invoiceNumber: string) =>
+          isMobile
+            ? wizard.mobileUsageMetaText(vendorName, invoiceNumber)
+            : wizard.usageMetaText(vendorName, invoiceNumber);
+        const containerOf = (invoiceNumber: string) =>
+          isMobile
+            ? wizard.mobileCard(vendorName, invoiceNumber)
+            : wizard.contentTableRow(vendorName, invoiceNumber);
+        const usageOf = (invoiceNumber: string) =>
+          isMobile
+            ? wizard.mobileUsageField(vendorName, invoiceNumber)
+            : wizard.usageField(vendorName, invoiceNumber);
+
+        // Area only → the leaf area name alone, as read-only meta text.
+        const areaOnlyMeta = metaOf(invoiceWithArea.invoiceNumber!);
+        await expect(areaOnlyMeta).toBeVisible();
+        await expect(areaOnlyMeta).toHaveText(areaName);
+
+        // Area + linked document → both halves, in that order, joined by " · ".
+        const bothMeta = metaOf(invoiceWithBoth.invoiceNumber!);
+        await expect(bothMeta).toBeVisible();
+        await expect(bothMeta).toHaveText(`${areaName} · 1 attachment: Invoice`);
+
+        // Neither → no meta element at all (not an empty one, no reserved gap).
+        await expect(metaOf(invoiceNoArea.invoiceNumber!)).toHaveCount(0);
+
+        // The meta text is READ-ONLY: even the row that has an attachments note exposes exactly
+        // ONE textbox (its Usage field) — the `Attachments note for …` `EditableField` is gone.
+        // Paired with the positive above, which proves the note is genuinely present, so this
+        // cannot pass just because the row has no attachment.
+        await expect(containerOf(invoiceWithBoth.invoiceNumber!).getByRole('textbox')).toHaveCount(
+          1,
+        );
+
+        // Neither half of the meta line is folded into the EDITABLE usage text (AC5.3) — that
+        // separation is what lets it survive manual edits and AI generation (Scenario 8 in
+        // `reportWizardAiGeneration.spec.ts`).
+        const usageValue = await usageOf(invoiceWithBoth.invoiceNumber!).inputValue();
+        expect(usageValue).not.toBe('');
+        expect(usageValue).not.toContain(areaName);
+        expect(usageValue).not.toContain('attachment');
+
+        // ...and there is no Attachments Note column header left on the desktop table (the
+        // mobile cards have no header row at all, so this half is desktop/tablet only).
+        if (!isMobile) {
+          const headerRow = wizard.contentTable.locator('thead tr');
+          await expect(headerRow).toContainText('Usage');
+          await expect(headerRow).not.toContainText(/attachment/i);
         }
       } finally {
         if (workItemWithAreaId) await deleteWorkItemViaApi(page, workItemWithAreaId);

From d48f0cdcde20135aa68b1058c02fd8dc5692d6d6 Mon Sep 17 00:00:00 2001
From: Frank Steiler <frank@steiler.dev>
Date: Mon, 3 Aug 2026 18:35:51 +0200
Subject: [PATCH 3/6] docs(memory): record PR #1959 review findings and the
 dontBreakRows bound rule

- Usage-cell height bound: bound what a cell RENDERS, never a source field
- Generic-assertion / hand-enumerated-input asymmetry as the recurrence cause
- Column-count constraint pinning MAX_SAFE_USAGE_CHUNK_CHARS to the 7-col shape
- Owed ADR-034 corrections (minimum-bar rule #1, module table, override keys)
- Two new recurring patterns: dead capability retained, AC reversal by polish issue

Co-Authored-By: Claude product-architect <noreply@anthropic.com>
---
 .../agent-memory/product-architect/MEMORY.md  |  4 +-
 .../product-architect/client-pdf-pipeline.md  | 46 +++++++++++++++++++
 .../product-architect/recurring-patterns.md   | 29 ++++++++++++
 3 files changed, 77 insertions(+), 2 deletions(-)

diff --git a/.claude/agent-memory/product-architect/MEMORY.md b/.claude/agent-memory/product-architect/MEMORY.md
index c966f33e6..373f93876 100644
--- a/.claude/agent-memory/product-architect/MEMORY.md
+++ b/.claude/agent-memory/product-architect/MEMORY.md
@@ -2,11 +2,11 @@
 
 ## Topic Files
 
-- [Recurring patterns & traps](recurring-patterns.md) — polymorphic FK cleanup, XOR CHECK vs SET NULL, forked-function drift, test smells, cross-layer contract drift, ajv `anyOf`, N+1 sites, async writes surviving state resets, cross-reference rot in documented-bound comments (#1939), usePreferences per-instance store + serialized-write-queue review (#1955)
+- [Recurring patterns & traps](recurring-patterns.md) — polymorphic FK cleanup, XOR CHECK vs SET NULL, forked-function drift, test smells, cross-layer contract drift, ajv `anyOf`, N+1 sites, async writes surviving state resets, cross-reference rot in documented-bound comments (#1939), usePreferences per-instance store + serialized-write-queue review (#1955), capability-retained-but-producer-removed (#1959), AC reversal by a polish issue (#1959)
 - [Dual-rail aggregation](dual-rail-aggregation.md) — Rail A/B tagged-deposit invariants (#1891/PR #1894), residual-denominator rule, isSplit UNION
 - [Source-report split inference](source-report-split-inference.md) — budgetLines[]/deposits[] are this-source-scoped, so †/‡ classification is a proxy; proposed `splitKind`; pdfmake `'2*'` width trap
 - [Story reviews](story-reviews.md) — per-story and per-PR review log
-- [Client PDF pipeline](client-pdf-pipeline.md) — ADR-034 report PDF generation + the reportContent content/layout split (#1900)
+- [Client PDF pipeline](client-pdf-pipeline.md) — ADR-034 report PDF generation, reportContent content/layout split (#1900), `dontBreakRows` silent-drop rule + **owed ADR-034 corrections** (#1959)
 - [Diary drafts pattern](diary-drafts-pattern.md) — ADR-022 draft lifecycle via status column on parent table
 - [EPIC-03 refinement](epic03-refinement.md) — 40 consolidated refinement items
 - [EPIC-04 household items](epic04-household-items.md) · [EPIC-05 budget](epic05-budget.md) · [EPIC-17 i18n](epic17-i18n.md) · [EPIC-18 areas & trades](epic18-areas-trades.md)
diff --git a/.claude/agent-memory/product-architect/client-pdf-pipeline.md b/.claude/agent-memory/product-architect/client-pdf-pipeline.md
index 4b9ee56de..746bccc37 100644
--- a/.claude/agent-memory/product-architect/client-pdf-pipeline.md
+++ b/.claude/agent-memory/product-architect/client-pdf-pipeline.md
@@ -215,3 +215,49 @@ baseline rather than blanking a field. Keep this shape — putting AI output in
 Server side: `POST /api/source-reports/generate-content` re-fetches the report (client sends selection only),
 persists nothing, and reuses the single `budgetExtraction/` LLM gateway as a third provider method. Both
 documented in ADR-034 "Addendum: content layers".
+
+### PR #1959 review — the Usage-cell height bound, third scope revision
+
+**The recurring defect** (#1929 r3 -> r4 -> #1959 r1) is always the same: `table.dontBreakRows: true` makes
+pdfmake measure an over-tall row and then `PageElementWriter` **silently discards** the overflow. No throw,
+no visible truncation. Symptom: page count **saturates** (~2) and can go **non-monotonic** (3 -> 2 as content
+grows) while rendered line count rises linearly.
+
+**The rule that fixes it durably: bound the height of what a cell RENDERS, never of a source field.**
+Any per-field bound is coupled to a layout decision and detaches silently the moment the layout changes —
+which is exactly how r4's per-field continuation rows became #1959's unbounded inline run.
+`packUsageCellRows(segments, MAX_SAFE_USAGE_CHUNK_CHARS)` is the right shape because the bound now lives
+where the cell's content **stream** is assembled, so a new segment is bounded automatically.
+
+**Two detection recipes that actually caught it** (both threshold-free, both worth reusing for any generated
+document): page count must be **monotonic** in content size, and **channel-independent** — the same text
+costs the same paper whichever field carries it.
+
+**The generic-assertion / hand-enumerated-input asymmetry** (the reason it recurred, now the top open item):
+the per-row budget assertion counts characters generically over all runs, so a new channel *would* be
+counted — but the test inputs are hand-listed (`usageText`/`areaText`/`attachmentsNote`), so a new
+`ReportContentRow` string field defaults empty and every assertion passes **vacuously**. Fix is key-driven
+saturation (`Object.entries(row)` -> saturate every string field), not another hand-written case.
+Verified empirically during review: losslessness + per-row budget hold over 3k fuzzed inputs (0 failures);
+a hypothetical **second** grey/meta segment trips the existing `splitUsageCell` "at most one grey run per row"
+throw in ~49% of inputs, so *that* channel class is already guarded.
+
+**Geometry constraint (blocks a feature):** `USAGE_WIDTH_7COL/_6COL` derive from `usableColumnWidth(n)`, and
+`MAX_SAFE_USAGE_CHUNK_CHARS = 650` was **measured against the 7-column shape**. So making column visibility
+affect the PDF widens the Usage column and **invalidates the measured ceiling** — "just plumb the toggles
+through" is a re-measurement story, not a UI change. #1959's toggles are preview-only local `useState`.
+
+**`packUsageCellRows` hangs on `maxChars <= 0`** (own `remaining <= 0 -> flush() no-op -> continue` loop).
+New to the packer — `splitIntoPageSafeChunks` fails loudly instead (`RangeError`). Unreachable while the
+budget is a constant; matters if it ever becomes computed.
+
+### ADR-034 debt (owed, NOT yet written — carry this forward)
+
+1. Add the `dontBreakRows` lesson + the "bound the rendered cell, not a field" rule + both detection recipes.
+2. **Minimum-bar rule #1 is wrong**: `table._minWidth <= 515.28` fails on correct code (`_minWidth` is the
+   widest unbreakable *word*, not the laid-out width). Correct check: `max(horizontalRatio) <= 1`.
+   B2's narrative is fine; the generalized rule was mis-transcribed.
+3. Module table drifted twice: add `pageGeometry.ts` (#1939) and `index.ts`; drop "PDF-local formatters" from
+   `shared.ts` (deleted in review round 2) and move "table layout constants" to `pageGeometry.ts`.
+4. Override-key list (line 148): drop `attachmentsNote` — unreachable since #1959.
+5. Record the fixed 6-or-7 column-count constraint above.
diff --git a/.claude/agent-memory/product-architect/recurring-patterns.md b/.claude/agent-memory/product-architect/recurring-patterns.md
index 2016c09f5..65a674ab1 100644
--- a/.claude/agent-memory/product-architect/recurring-patterns.md
+++ b/.claude/agent-memory/product-architect/recurring-patterns.md
@@ -389,3 +389,32 @@ in flight. That is the whole proof. Two corollaries worth asserting in review:
 Ordering claims: a client queue genuinely fixes server-side out-of-order application only if the next
 write is not _dispatched_ until the previous response is back (i.e. `await` wraps the HTTP call, not just
 the local state update). Check that, or the fix only narrows the window.
+
+## Capability retained in code, disabled at the producer, documented by comment (PR #1959)
+
+Three instances in one PR, all the same shape — **a type permits a state the producer can never emit, and the
+constraint is enforced by a comment instead of by the type**:
+
+- `ReportContent.footnotes` is now unconditionally `[]`, yet 4 consumers still branch on / copy it. The E2E
+  page object had to add a prose warning ("never assert a positive count on these").
+- `applyOverrides` still honours `row.<id>.attachmentsNote` with **no producer left** (editor field removed;
+  overrides are ephemeral wizard state — no server persistence, nothing in migrations 0001-0044; and
+  `applyAiContent` only writes `usageText` + cover-letter fields).
+- Column-visibility toggles labelled `Show/hide columns` that affect the preview but **not** the exported PDF.
+
+Rule: **delete rather than comment.** A write path with no reader and no producer is how the #1929
+round-3/round-4 confusion started — the code said one thing and the comment said another. When triaging a
+"keep it as a capability?" question, check for a *producer* first: no producer => dead, remove it.
+
+## Verify the AC record when a PR reverses a recently-shipped story (PR #1959)
+
+#1959 reversed #1923's AC1.1/1.2/2.3/2.4 **one day** after #1923 shipped (2.13.0-beta.38, 2026-08-02), while
+#1923 stayed CLOSED with those ACs marked delivered — so the record asserted two contradictory behaviours were
+both correct on `beta`, and the pending promotion PR would have summarized reversed ACs as delivered.
+
+Tell: the reversing issue had **zero comments**, no `**[product-owner]**` header, no numbered ACs (its body was
+a PR-style summary), and no ux-designer visual spec — whereas the story it reversed had all four. **A
+requirements reversal authored as a polish issue is the signature.** Cheap fix: PO supersession comment on the
+old issue + PO ratification on the new one. Check this whenever a PR deletes user-visible report/document
+content — and check whether the replacement text preserves *meaning* (`(abzgl. Abschlag)` lost the footnote's
+"claimed separately", which is compliance-relevant in a bank-facing document).

From de422091defbf87a92ccf8266633f87f87d6d63d Mon Sep 17 00:00:00 2001
From: Frank Steiler <frank@steiler.dev>
Date: Mon, 3 Aug 2026 18:39:29 +0200
Subject: [PATCH 4/6] docs(memory): record PR #1959 product rulings and
 glossary decisions

- Footnote-to-inline-label reversal: labels kept, explanatory sentences
  restored as a report-level legend (#1965); supersession-comment policy
  for ACs on closed/released issues instead of rewriting them
- Glossary: Abschlag approved as a measured-space short form; split's
  three German surface forms deliberately get no entry

Co-Authored-By: Claude product-owner <noreply@anthropic.com>
---
 .claude/agent-memory/product-owner/MEMORY.md  |  1 +
 .../product-owner/bank-report-wizard.md       | 47 +++++++++++++++++++
 2 files changed, 48 insertions(+)

diff --git a/.claude/agent-memory/product-owner/MEMORY.md b/.claude/agent-memory/product-owner/MEMORY.md
index d6c6a9623..7ebea615a 100644
--- a/.claude/agent-memory/product-owner/MEMORY.md
+++ b/.claude/agent-memory/product-owner/MEMORY.md
@@ -41,6 +41,7 @@ Full detail in [standalone-bugs-and-stories.md](standalone-bugs-and-stories.md)
 - **DataTable: #1955** two-column toggle race silently hides 2nd column, all 6 DataTable pages (Should Have, S, Backlog, 2026-08-02). See [datatable-column-preference-race.md](datatable-column-preference-race.md) — records that **fast clicking is the SAFE case** (I judged this backwards; debounce `clearTimeout` coalesces rapid input, the >500ms reading-pace gap is the reachable one) and that #1920's E2E-only fix (`InvoicesPage.enableColumn()` awaits the PATCH) makes CI green **without** fixing production — don't close #1955 on a green shard.
 - **#1957** latent cross-file E2E test-isolation hazard (shared-admin `user_preferences` writes under `fullyParallel`, `LocaleContext.syncWithServer` actively flips a victim test's locale) — Should Have, bug, Backlog, 2026-08-02/03, filed from `/fix-e2e` work on PR #1956. Scoped as an audit + per-spec sweep, not a single-file fix — found a second live instance (`diary-uat-fixes.spec.ts` vs `dashboard.spec.ts`, key `dashboard.hiddenCards`) while researching it. Distinct from #1955 (production race) and #1920 (E2E workaround for #1955). Detail in [e2e-shared-admin-preference-hazard.md](e2e-shared-admin-preference-hazard.md).
 - **Bank Report Wizard mini-epic** (no parent epic): #1876 refunds (PR #1880) → #1877 contact/household/attachment typing (PR #1883) → #1878 report backend → #1879 wizard+PDF (PR #1887, CHANGES_REQUIRED then **APPROVED** round 2, 2026-07-30). Domain decisions, contract facts (refund sign, `isSplit`, `totalAmount` scope) and deferrals in [bank-report-wizard.md](bank-report-wizard.md). **Refinement Round 2** (2026-07-31, no parent epic, all Todo): #1898 report table refinements (PR #1902, merged) → #1899 settings step + report language (PR #1903, PO review 2026-07-31: **APPROVED w/ 1 MUST FIX** — 5-step wizard, `getFixedT`/`createFormatters` threading and en/de report output all verified; AC 2.2 default-locale seeding is stale on hard load, see [pr-review-patterns.md](pr-review-patterns.md) `useState(contextValue)` entry) → #1900 editable HTML preview (PR #1909, round 1 COMMENT + 4 MUST FIX, **round 2 2026-07-31 APPROVED** — all 4 verified on disk: local `composes` classes, `--font-weight-medium`, `sharedStyles.srOnly`, translated `resetFieldAriaLabel` at all 9 sites w/ en+de parity; stylelint exit 0. Note: `gh pr review --approve` fails when PO authored the PR context — post verdict via `gh pr comment` with explicit Verdict line. Judgment rulings: signature-derived-from-sender ACCEPTED, mark-claimed-generates-no-PDF ACCEPTED as vacuous, per-field reset ACCEPTED, AC 4.6 rendered-preview assertion ACCEPTED as documented deviation — Playwright headless has no PDF viewer plugin, so the E2E asserts the CSP `frame-src` contract instead; **mixed-language mobile cards ACCEPTED** — see [bank-report-wizard.md](bank-report-wizard.md) "artifact content vs. edit affordance") → #1901 AI usage/cover-letter generation (PR #1916, PO review 2026-07-31: **CHANGES_REQUIRED** — 3 blocking numeric-accuracy defects in the LLM prompt inputs: `/100` on major-unit amounts, `Math.round` to whole euros, per-invoice amount ignoring `excludedLineIds`; + 2 MUST FIX: extraction-flavoured shared LLM error copy, uncommitted wiki API-Contract section. All 6 AC sections otherwise met; entity-level linked-item description deviation ACCEPTED. New defect class recorded in [pr-review-patterns.md](pr-review-patterns.md) "LLM/prompt-assembly defects". **Round 2 on `b70d821b`: APPROVED** — all 5 findings fixed and verified on disk; `prompts.test.ts` gained a dedicated ×100 regression-guard block (98/98 pass locally); per-invoice cents-rounding now makes server math identical to client `applyLineExclusions`; wiki pushed at `254db1d`; the 9 removed test lines were a stale #1915 header note, not a weakened assertion). **Follow-ups consolidated into #1917** (tech-debt, Should Have, Backlog): architect M1–M4 + L1/L2/L3/L5, the `Konstruktionsprojekt`→`Bauprojekt` prompt nit, and the approved `KI` glossary entry. M2 (extract `computeIncludedTotal` to `@cornerstone/shared`) is the headline — the client/server duplication already drifted once and caused the #1916 blocking bug. Open: **#1891** user-verification follow-up (Todo, PR #1894 **APPROVED** 32/32 round 2, 2026-07-30 — 2 wiki MUST FIX outstanding); **#1888** stage-matched attachment indicator (Backlog, blocked-by #1879); **#1895** HIGH claim close-out cross-source sweep, **#1896** quotation-deposit 409 (blocked-by #1895), **#1897** deposit-blind drill-down — all Backlog, from the #1891 architect audit; **#1910** `lang` attribute on report-language preview content (Backlog, a11y follow-up from #1909 round 2); E2E shard 5 pre-existing flake must be triaged before promoting to `main`. **Refinement Round 3** (2026-08-02, from user PDF inspection + wizard walkthrough, all Todo, for `/batch-develop`): **#1929** PDF layout robustness (bug, Must Have — column widths, `dontBreakRows`, header clipped by 40pt top margin; **PR #1935 CHANGES_REQUIRED ×2, AC2-vs-AC4 conflict ruled 2026-08-02: precedence ladder I1 no-loss > I2 no-clip > I3 row-whole > I4 no-word-break; AC2/3/4 rewritten, AC12–AC14 added; 600-char target**), **#1930** attachment tier rules per report type (quotation→deposit→invoice; null = tier `invoice`; supersedes #1888's design question) — **PR #1942 APPROVED round 1, 2026-08-02**, all 11 AC met, 80/80 green; **but #1943** (bug, **Must Have**, Todo, 2026-08-02) — `handleUseCaseChange` never clears `report`/`sourceId`, so budget-overview→claim carries a stale report and can embed **quotation-tier docs in a claim PDF**, reaching #1930 AC2's forbidden outcome by a route AC2 doesn't cover; ruled: clear `sourceId` too — **PR #1942 APPROVED round 1, 2026-08-02**, all 11 AC met, 80/80 green; **#1888 body re-scoped to indicator presentation only at review time** (it was still stale), **#1931** single "Enhance with AI" button + purpose-focused prompt (takes the `Konstruktionsprojekt` nit off #1917), **#1932** cover letter overhaul (folds in #1925, reverses #1909's derived-signature acceptance) — **PR #1951 APPROVED round 1, 2026-08-02**, all 40 ACs met incl. both struck-as-vacuous negative constraints; 3 MUST FIX (German `Mit freundlichen Grüßen` comma, unpinned Closing row, #1925 AC5 editor pin). **#1925 CLOSED as duplicate** (board Wont-Do) — my §6 carried 4 of its 6 ACs, a second instance of the AC-transcription failure mode, **#1933** Select Invoices step UI fixes. Rulings in [bank-report-wizard.md](bank-report-wizard.md) §"Refinement Round 3". **PR #1945 (#1943) review follow-ups filed 2026-08-02**: **#1946** in-flight AI generation survives a use-case change (bug, **Must Have**, Todo — product ruling: widen `guardedUpdate`'s dirty predicate to include `isGeneratingAi`, confirm invalidates via token, cancel lets it finish) and **#1947** `ReportWizardPage` `useReducer` refactor (tech-debt, Should Have, Backlog, blocked-by #1946, filed separately from #1912 on purpose). **M1 was fixed inside PR #1945, not filed** — precedent: a finding that defeats an AC of the story under review belongs in that story's PR regardless of reviewer severity. #1943 AC4 reworded + AC5 enumeration completed (`skippedDocuments`/`aiError` ruled CLEAR, carried as #1946 AC9/AC10); #1933 AC2.1/2.7 corrected (invoice row has no mobile card). See [bank-report-wizard.md](bank-report-wizard.md) §"#1945 review follow-ups" and §"#1933 ACs 2.1/2.7" — the latter records a **recurring AC-writing failure mode** (ACs that misdescribe reality fail correct implementations at UAT). **#1929 CLOSED 2026-08-02** — PR #1935 merged (squash `1c5aa62c`) after **4 rounds**; both reviewers measured by real render+rasterize. 5 follow-ups filed: **#1937** German header labels break mid-word (bug, Todo, translator fast-follow — widening measured and rejected), **#1938** running-header `generated at` label with no timestamp on pages 2+ (bug, Todo, **pre-existing**), **#1939** reportPdf geometry hygiene (tech-debt, Todo, **blocks #1932** — `HEADER_ROW_HEIGHT`→`_MAX` 68pt vs measured 45.81pt, char-advance comment scoping, `PDF_STYLES` relocation), **#1940** continuation rows read as broken (could have, Backlog), **#1941** override fields have no `maxLength` (could have, Backlog), **#1950** guard test recomputing the derived `Ѹ` ceiling (tech-debt, could have, Backlog, blocked-by #1939 — filed 2026-08-02 from PR #1948 §2; architect reframed its own ask from "re-run the 3,919-codepoint sweep" to "a test that recomputes 616 from `USAGE_WIDTH_7COL`/font sizes/`DEFAULT_LINE_HEIGHT`", sweep now an explicit non-goal. Rulings: **comment keeps the rationale, issue owns the guard, neither replaces the other**; bounded-and-quantified earns a tracked owner where unbounded-and-estimated gets documentation only; a derived bound with no test is a comment waiting to go stale — `overviewPdf.test.ts` pins `704`/`546` as literals tied to no geometry constant). `markerText`+`invoiceNumber` folded into #1939 as documentation-only; vendor-name mid-word break recorded as accepted limitation in #1937. Detail in [bank-report-wizard.md](bank-report-wizard.md) §"#1929 closed". **#1931 PR #1944 APPROVED round 1, 2026-08-02** — all ACs met **except 3.2/3.3, deliberately NOT claimed**: they assert live-model output quality, which a mocked LLM cannot verify. Ruling: **merge is a code gate, Done is an acceptance gate** — PR merges, story stays out of Done until a human reads real EN+DE output with `LLM_*` set; UAT scenarios posted on #1931; failure → reopen #1931, don't file a follow-up. Contrast #1909 AC 4.6: an unverifiable AC **with** a substitute assertion may be waived as a documented deviation; **without** one it goes to UAT. "Mit KI verbessern" accepted for AC 2.3. **#1917 L3 struck** (verified fixed); rest of #1917 open, **`KI` glossary entry still #1917's** — `glossary.json` untouched by #1944. Detail in [bank-report-wizard.md](bank-report-wizard.md) §"#1931 reviewed".
+- **PR #1959 (report PDF UX, user's own PR, held promotion #1958) — 3 rulings 2026-08-03.** Footnotes `†`/`‡` + their sentences replaced by inline labels `(partial)`/`(less deposit)`, reversing **#1923 AC1.1/1.2/2.3/2.4** (which had already superseded #1898 §4). **Ruled: labels kept, sentences return as a non-editable report-level legend → #1965 (Must Have)** — `(partial)` is near-self-evident beside the two amount columns, but `(less deposit)` drops "claimed **separately**", a materially different audit claim. Cheap because the footnote channel is orphaned-not-removed (producer-only fix). **Closed/released ACs get a dated supersession comment, never a rewrite** (posted on #1898 + #1923). #1923 AC5.3 substance survives — `areaText` still a separate field, so judged against its stated rationale, not its prescribed sub-line rendering. Glossary: **`Abschlag` APPROVED** as a measured-space short form of `Abschlagszahlung` (option (c) is arithmetically impossible — the full term eats 72.85 of 75pt, so no qualifier fits; collapsing into the constituted-deposit label would lose more); **`split`'s three German forms get NO entry** (glossary prevents semantic divergence, is not a surface-form registry). Both on **#1917**. Issues filed: **#1965** legend, **#1966** column-toggle E2E, **#1967** `attachmentsNote` override unreachable, **#1968** meta-suffix single run, **#1969** `testPrefix`/`authenticatedPage`, **#1970** configurable auth rate limits, **#1971** `search-users.spec.ts` leftovers, **#1972** silent failed column-pref saves + dead `isLoaded`. Full rulings in [bank-report-wizard.md](bank-report-wizard.md) §"PR #1959 rulings".
 
 ## Requirements Coverage
 
diff --git a/.claude/agent-memory/product-owner/bank-report-wizard.md b/.claude/agent-memory/product-owner/bank-report-wizard.md
index d7fd7ce0d..a0596a427 100644
--- a/.claude/agent-memory/product-owner/bank-report-wizard.md
+++ b/.claude/agent-memory/product-owner/bank-report-wizard.md
@@ -396,3 +396,50 @@ Architect's PR #1951 review ([comment 5160566459](https://github.com/steilerDev/
 ### PR body staleness — checked downstream, clean
 
 The "No new E2E coverage — deferred" line was already corrected (now lists Scenarios 21-24, checkbox ticked). Checked both propagation paths: no open `beta → main` promotion PR, and `RELEASE_SUMMARY.md` predates the report-wizard work entirely (it covers Cost Basis filter / auto-itemize merge / diary defaults) and is regenerated by `docs-writer` at promotion. Nothing downstream consumed the stale claim.
+
+## PR #1959 rulings (2026-08-03) — footnotes removed, glossary short-form approved
+
+`fix(reports): improve report PDF UX`, the **user's own PR**, held the `beta → main` promotion (#1958). Three product questions, all ruled without blocking promotion.
+
+### The reversal: shared footnotes → inline labels
+
+#1959 removed the `†`/`‡` markers **and their explanatory sentences**, replacing them with inline grey labels on the allocated amount: `(partial)`/`(Teilbetrag)` and `(less deposit)`/`(abzgl. Abschlag)`. This reverses **#1923 AC1.1, AC1.2, AC2.3, AC2.4** — which #1898 §4 had itself already been superseded by. `allocatedMarkers` gone from `types.ts`, replaced by `isSplit`/`isDepositReduced`.
+
+**Ruling: labels accepted, sentences must come back as a report-level legend (#1965, Must Have).** The differentiator is *which* sentence is load-bearing:
+
+- `(partial)` is nearly self-evident — the table prints **Invoice Amount** and **Allocated Amount** side by side, so the label only has to name the reason for a difference the reader can already see.
+- `(less deposit)` loses a **materially different claim**. The footnote said the deposit was claimed **separately** — accounted for in another submission. "less deposit" is equally consistent with the deposit never being claimed, and leaves a double-claim question open when it resurfaces. On a Verwendungsnachweis/Mittelabruf that has audit consequences.
+
+**Generalizable: for outbound financial copy, ask what a label lets the reader *conclude*, not whether it is accurate. `(less deposit)` is true and still misleading by omission.**
+
+**Why the fix is cheap (checked, load-bearing for the ruling):** the footnote *rendering channel is fully intact* and merely unused. `buildReportContent.ts` declares `footnotes: ReportContentFootnote[] = []` (L232) and returns it (L289) but never pushes; `ReportContentEditor.tsx` L445-450 renders it; `overviewPdf.ts` appends `reportContent.footnotes` **verbatim** (pinned by the `overviewPdf.test.ts` case "appends reportContent.footnotes verbatim…", which is also the only surviving reader of the two i18n keys — as fixture text). So the legend is a **producer-only change** and preview/PDF parity comes free. **Pattern: before ruling a restoration too expensive, check whether the mechanism was removed or only orphaned.**
+
+**Legend must NOT go in the cover letter** — it is user-editable (#1932), so an editable qualification can be deleted, silently removing a material statement from a financial document. Report-level and non-editable, like the existing footnote block.
+
+**The orphan-key cleanup was deliberately NOT filed as a deletion issue.** `splitFootnote`/`depositReducedFootnote` are *consumed* by #1965, so filing "delete these orphans" would race the legend. #1965 AC 3.2 pins retention; its Notes record deletion as the Wont-Do alternative. **Pattern: when a cleanup follow-up and a restoration follow-up target the same artifact, one issue must own both or the cleanup wins by arriving first.**
+
+### What was NOT an AC reversal (checked, initially framed as one)
+
+**#1923 AC5.3 substance survives.** The area name moved from a sub-line to an inline grey suffix, but `areaText` is **still a separate row field** (`buildReportContent.ts` L196/L214) and `applyAiContent.ts` still only assigns `row.usageText` (L50) — so AI-generated usage text cannot drop the area, which was AC5.3's *stated rationale*. #1923's own Notes delegated the area sub-line's **visual treatment** to `ux-designer`. So E2E Scenario 20's rewrite is a presentation change within delegated authority. **Pattern: an AC that states its own rationale should be judged against the rationale, not the prescribed rendering — that is what the rationale is there for.**
+
+### Stale ACs on closed/released issues — comment, don't rewrite
+
+Both #1898 and #1923 are CLOSED + `released on @beta`. **Ruled: do not rewrite their ACs; post a dated supersession comment** naming which ACs died, by which PR, what still holds, and where the new source of truth is. Rewriting a released story's ACs falsifies the record of what was accepted and loses the design reasoning. The standing "stale AC is a real defect" rule targets ACs that a *future* implementer or UAT run will read as live spec — a supersession comment discharges that risk without destroying history. Posted on both (#1923 comment also lists the non-superseded ACs explicitly, since AC2.1/2.2/2.5/3/4 all still hold).
+
+Also noted on #1898: its §4 has now been superseded **twice** (#1898 → #1923 → #1959). That churn is itself the argument that the footnote *presentation* was never settled, and it supports abandoning the glyphs — which is why #1965 keeps the sentences but not the markers.
+
+### Glossary: `Abschlag` APPROVED as a space-constrained short form
+
+Full ruling on **#1917** (the pending glossary-refinement pass, same place the `KI` entry lives; translator implements, PO does not edit `glossary.json`).
+
+`depositReducedInlineLabel` = `abzgl. Abschlag` ships as written, against the glossary-approved `Deposit → Abschlagszahlung`. Translator measured with fontkit against the real embedded `Roboto-Regular.ttf` at 8pt in the fixed 75pt `ALLOCATED_AMOUNT_WIDTH`: `" (Abschlagszahlung)"` = 72.85pt (the existing sibling badge, so it **sets the real ceiling**), `" (abzgl. Abschlag)"` = 63.95pt, `" (abzgl. Abschlagszahlung)"` = **96.07pt, overflows by ~21pt**.
+
+**The decisive arithmetic the options list missed: option (c), "a shorter compliant string", is unavailable.** `Abschlagszahlung` alone eats 72.85 of 75pt, so **no qualifier of any length fits** — not `abzgl.`, not `ohne`, not the accounting `./.`. Keeping the full term therefore forces dropping the qualifier, collapsing this label into the *constituted*-deposit label (`(Abschlagszahlung)`, #1923 AC2.1). Those are different facts — "this row **is** a deposit" vs "this amount is **reduced because** deposits were claimed separately". **Collapsing them is a worse information loss than the abbreviation.**
+
+Option (b), widening the column, rejected on **risk not cost**: `ALLOCATED_AMOUNT_WIDTH` is #1929 geometry that took **four rounds** and real render-and-rasterize measurement, precisely because Usage was collapsing and the table overflowed the page edge. Reopening it for zero reader benefit is a bad trade. And `Abschlag`/`Abschlagszahlung` are the same concept in German construction practice (cf. `Abschlagsrechnung`); `abzgl. Abschlag` is idiomatic invoice German. No reader is misled — the bar set by the Verwendungsnachweis/Einreichung precedent.
+
+**Why it earns a glossary entry rather than a silent exception: without one a future compliance sweep "fixes" it in good faith and silently breaks a bank-facing PDF, invisibly to the unit suite.** The entry must record the 75pt column and the measurement as the *reason*, not just list the variant. **Generalizable: a deliberate deviation from an approved term needs a recorded reason at the glossary, or the next sweep reverts it.**
+
+### Glossary: `split`'s three German forms — NO entry, deliberately
+
+`anteilig` (adj.) / `Anteil` (noun) / `Teilbetrag` (noun), each role-correct, mirroring English's own `partial`/`split`/`portion`. **Ruled: not drift, no entry.** The glossary prevents *semantic* divergence — one concept becoming two concepts. It is **not a single-surface-form registry**, and pinning one form here would force ungrammatical copy across an adjective and two nouns. Recorded on #1917 as "reject if proposed later" so a future translator does not re-escalate. Revisit only if the *English* is unified — a copy story, not a glossary one.

From fc5db9613427a6bfdd54f5b6218b005a39f30ef8 Mon Sep 17 00:00:00 2001
From: Frank Steiler <frank@steiler.dev>
Date: Mon, 3 Aug 2026 18:39:47 +0200
Subject: [PATCH 5/6] docs(memory): record #1959 inline-label wrap finding and
 review patterns

- Multi-word inline labels split mid-phrase in narrow fixed-width PDF
  cells; an isolated width measurement cannot catch it. Verified U+00A0 fix.
- Check that a new report/export editor control actually reaches the
  export pipeline; DataTableColumnSettings already exists for column toggles.
- Compare preview-component styles against the exporter's constants.

Co-Authored-By: Claude ux-designer <noreply@anthropic.com>
---
 .claude/agent-memory/ux-designer/MEMORY.md    |  3 +++
 .../ux-designer/pr-review-findings.md         | 22 +++++++++++++++++++
 2 files changed, 25 insertions(+)

diff --git a/.claude/agent-memory/ux-designer/MEMORY.md b/.claude/agent-memory/ux-designer/MEMORY.md
index a95b754a5..754bfa7b8 100644
--- a/.claude/agent-memory/ux-designer/MEMORY.md
+++ b/.claude/agent-memory/ux-designer/MEMORY.md
@@ -36,3 +36,6 @@
 - A component that renders its own internal `<label><input/></label>` (e.g. `TriStateCheckbox`), when wrapped in an *outer* `<label>` alongside other row content, forwards clicks anywhere in that outer box to the checkbox. Never place a new interactive affordance inside that label — give it a sibling grid cell instead (structural guarantee beats `stopPropagation`). See Issue #1933 in [feature-spec-history.md](feature-spec-history.md).
 - `display: flex; flex-direction: column` on a `<td>` desyncs it from sibling cells' baseline alignment (synthesized flex-baseline ≠ real text baseline), independent of font-size. Fix with `vertical-align: top` on the shared `td` rule plus dropping flex from the desktop cell (replace `gap` with `margin-top` on non-first children); where the same class is reused as a mobile `<div>`, keep flex+gap there under the mobile media query. See Issue #1933 in [feature-spec-history.md](feature-spec-history.md).
 - `.expandButton`-style `padding:10px; margin:-10px` (small visible box, 44px hit target via negative margin into existing gutter) is this codebase's house technique for a real touch target on a small row icon — prefer it over a literally-44px visible box.
+- **Multi-word inline labels split mid-phrase in narrow fixed-width PDF cells** — a width measurement of an appended annotation *in isolation* proves nothing; it always shares the line with the value it annotates, so measure the combination, and check whether the string can break internally at all. Verified fix: U+00A0 for the internal space (width bit-identical, no geometry change). Single-word labels wrap cleanly as a unit; only the two-word one broke, in BOTH locales. See PR #1959 in [pr-review-findings.md](pr-review-findings.md).
+- **A new control in a report/export editor may not reach the export pipeline** — grep its state symbol outside the component before assuming it does (PR #1959: column-visibility checkboxes changed only the on-screen preview, never the PDF). For column show/hide specifically, `DataTable/DataTableColumnSettings.tsx` already exists — check it before accepting a bespoke checkbox row.
+- When a component previews an exported artifact, compare each style against the *exporter's* constants, not only against design tokens — PR #1959 shipped two grey annotations at two sizes, only one matching the PDF.
diff --git a/.claude/agent-memory/ux-designer/pr-review-findings.md b/.claude/agent-memory/ux-designer/pr-review-findings.md
index a6b6a0c98..0db264382 100644
--- a/.claude/agent-memory/ux-designer/pr-review-findings.md
+++ b/.claude/agent-memory/ux-designer/pr-review-findings.md
@@ -123,3 +123,25 @@ Script-verify method for duplicate-key removals: parse the pre-PR JSON with Pyth
 Bonus finding (informational, not blocking, not caused by this PR): even after dedup, `schedule.json`'s surviving `detail.edit` object (`title`/`form.*`) and `diary.json`'s `create.title`/`edit.title` keys are themselves entirely unreferenced by any `t()` call — real page headings/field labels live under separate `createPage.*`/`editPage.*`/`entryForm.*`/`detail.view.*` keys. Orphan-key sweeps that target only _literal duplicates_ (this PR) or only _known-dead feature keys_ (the 5 photoAnnotator keys) can still leave a residue of never-referenced-at-all keys behind — worth a follow-up sweep that cross-references every namespace key against `grep -rn "t('<ns-prefix>"` call sites, not just against duplicate-key detection.
 
 For dead-CSS verification: CSS Modules are locally scoped, so a repo-wide grep for a removed class name (e.g. `modalActions`) will hit dozens of _unrelated_ components' own same-named local classes — always narrow the grep to the specific component file that imports the CSS module being edited, not the whole repo.
+
+## PR #1959 — report PDF UX (inline meta, inline split/deposit labels, cover-letter paragraphs, column toggles) — CHANGES_REQUIRED via `gh pr comment` (own PR), https://github.com/steilerDev/cornerstone/pull/1959#issuecomment-5169138685
+
+**The blocking finding, and the generalisable lesson: a multi-word inline label in a narrow fixed-width cell splits mid-phrase, and a width measurement of the label *in isolation* will never catch it.** #1959 replaced the non-wrapping single-glyph `†`/`‡` footnote markers with parenthesised inline labels appended to the 75pt right-aligned Allocated Amount cell. `depositReducedInlineLabel` (`less deposit` / `abzgl. Abschlag`) is the ONLY inline label in either locale containing an internal space — every other one (`partial`/`Teilbetrag`, `Deposit`/`Abschlagszahlung`, `(refund)`/`(Rückerstattung)`) is a single word and so wraps as an intact unit onto its own line. The two-word one wraps *at its own internal space*, rendering `(Teilbetrag) (abzgl.` / `Abschlag)` (de, combined case) and `€4,000.00 (less` / `deposit)` (en, isolated case) — opening and closing bracket on different lines. **Both locales**, on ordinary data.
+
+The PR's defence was a font measurement: `" (abzgl. Abschlagszahlung)"` = 96.07pt vs a 75pt column, so the shorter `" (abzgl. Abschlag)"` (63.95pt) was chosen "because it fits". That measured the label alone, but the label always follows the formatted amount in the same run stream (`4.000,00 €` ≈ 48pt), so the combined run *always* exceeds the column and the label *always* wraps. Width was never the operative variable — **breakability** was. Generalise: when reviewing any "we picked the shorter string so it fits column X" claim, ask what else shares the line, and whether the string can break internally. A width budget for an appended annotation is meaningless unless it's measured *including the value it appends to*.
+
+**Verified fix (rendered, not asserted): replace the internal space with U+00A0.** Width is bit-identical (NBSP has the same advance as a space in Roboto), no geometry constant changes, two string literals in `client/src/i18n/{en,de}/budget.json`. Rendered confirmation: label becomes one intact right-aligned line, exactly like `(Teilbetrag)` already behaved, in both the isolated and combined cases. Remember this as the go-to fix for any wrapped-parenthetical defect in these fixed-width PDF cells.
+
+**Column-visibility toggles: state never leaves the component.** `hiddenColumns` is `useState` local to `ReportContentEditor`, no prop/callback, and `grep` for `hiddenColumns|ColumnKey|columnVisibility` outside the file returns nothing — so the checkboxes change only the on-screen preview table; the generated PDF always contains every column. Issue text said "local state, no persistence needed", so intentional, but in a wizard whose sole output is a PDF and where every *other* editor control (`EditableField`) does change the PDF, "Show/hide columns" reads as "choose the report's columns". **Whenever a report/export editor gains a new control, check whether it actually reaches the export pipeline before assuming it does** — and if it doesn't, that's a finding even when deliberate.
+
+Also: `client/src/components/DataTable/DataTableColumnSettings.tsx` already is the app's column show/hide UI (gear trigger, checkbox list, reset-to-defaults, Escape + outside-click). #1959 built a bespoke inline checkbox row instead. Defensible on the merits (that component is bound to `ColumnDef<T>` + `DataTable.module.css` and is desktop-only, whereas these toggles must also drive the mobile card list) but still a Component Reuse Policy rule-3 violation: a new pattern must be extracted as a shared component, not left inline in a page component. Check `DataTableColumnSettings` first for any future column-visibility request.
+
+**Editor-vs-artifact fidelity is its own review axis here.** `.inlineNote` is `--font-size-xs` while the PDF renders the same annotation at `DEPOSIT_NOTE_FONT_SIZE` (8pt = body size); `.usageMetaText` meanwhile *dropped* the `font-size: xs` its predecessor `.usageAreaText` had and so now *does* match the PDF. Two grey annotations, two sizes, only one truthful to the artifact. When a component is a preview of an exported document, compare each style against the exporter's constants, not just against the design tokens.
+
+**Removing footnotes removes information asymmetrically.** `(partial)` stays self-evident because `Invoice Amount` and `Allocated Amount` sit side by side in adjacent columns — the removed `splitFootnote` sentence was redundant. `(less deposit)` does NOT: nothing on the page says what was deducted or where it was claimed, so dropping `depositReducedFootnote` is real information loss in a document that exists to justify a disbursement. Both keys survive unused in both locale files. Lesson: when a PR replaces explanatory footnotes with terse inline labels, evaluate each label separately for whether the surrounding columns already carry its meaning — don't rule on the batch.
+
+**`attachmentsNote` became unreachable-but-honoured.** Its `EditableField` was deleted with the column, so the `row.<id>.attachmentsNote` override key can no longer be produced from the UI, while `applyOverrides` still applies it — a pre-existing saved override renders with no edited-marker and no reset control. `types.ts` documents this candidly. Watch for this shape (override/preference key still consumed but no longer producible) whenever an editable field is demoted to static text.
+
+**Pre-existing, confirmed untouched by the diff, worth their own issues:** (1) `merge.ts:131` passes `t('sourceReports.table.generatedAt')` — bare `"Generated"` / `"Erstellt am"` — as `buildPageHeader`'s `generatedAtLabel` with **no date appended**, so every page after the first prints a dangling label trailing into nothing; (2) `VENDOR_SAFE_TOKEN_CHARS` resolves to 6 chars so essentially every real vendor name breaks mid-word with no hyphen (`Nordwind W`/`indows`, `Keller Concr`/`ete`) — known from #1929 round 4, re-confirmed; (3) a trailing blank page appeared in one of four fixtures (en claim, 9 rows) with the table well clear of the page-2 boundary — not root-caused, not attributed, but taller rows make boundary conditions likelier; (4) no `color-scheme` declared anywhere in `client/src/styles/`, so native checkboxes keep light browser chrome in dark mode (app-wide).
+
+**Fixture hygiene reminder:** my first render printed the raw key `sourceReports.table.attachmentType.certificate` into the PDF and I nearly filed it as a defect — the cause was my own fixture using `attachmentType: 'certificate'`, which is not in the `AttachmentType` union (`'quotation' | 'deposit' | 'invoice'`, `shared/src/types/document.ts:31`). **Before reporting an untranslated-key or missing-label finding from a hand-built fixture, check the fixture's enum values against the shared type union.**

From 87245a74a444e6d1e3913ad308476f4280b6d486 Mon Sep 17 00:00:00 2001
From: Frank Steiler <frank@steiler.dev>
Date: Mon, 3 Aug 2026 19:09:53 +0200
Subject: [PATCH 6/6] fix(reports): stop the deposit-reduced label splitting
 its own brackets
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

depositReducedInlineLabel was the only inline label in either locale with
an internal space, so at 8pt in the narrow allocated-amount column it
wrapped there — putting the opening and closing bracket on different
lines. "EUR 4,000.00 (less" / "deposit)" in en, "(Teilbetrag) (abzgl." /
"Abschlag)" in de, in a document sent to a bank. Both now use U+00A0,
which has the same glyph advance, so no width or geometry constant moves.

The earlier width analysis measured the label alone against the 75pt
column. It always follows the formatted amount in the same run stream, so
the combined run exceeds the column regardless and the label always
wrapped: breakability was the operative variable, not width. Verified by
rendering a plain-space control from the same harness and reproducing the
break, then confirming it gone.

Adds a locale-level invariant so this cannot regress silently: the two
keys that exist solely as bracketed labels must contain no breaking
whitespace, checked against every locale. attachmentType.deposit is held
to a weaker single-word rule instead, because it is also comma-joined
into flowing prose where a non-breaking space would be over-reach.

Also guards packUsageCellRows and splitIntoPageSafeChunks against a
non-positive budget. Both hung rather than failing loudly — the packer in
its own loop, before it ever delegates — so guarding only the primitive
would not have fixed it.

Labels the ReportContentEditor column toggles preview-only: their state
never leaves the component, but every sibling control does reach the PDF,
so the label read as "choose the report's columns". Wiring is #1966.

Glossary records Abschlag as a measured-space short form of
Abschlagszahlung, with the 75pt budget as the reason, so a future
compliance sweep cannot silently break the PDF.

Co-Authored-By: Claude frontend-developer <noreply@anthropic.com>
Co-Authored-By: Claude qa-integration-tester <noreply@anthropic.com>
Co-Authored-By: Claude translator <noreply@anthropic.com>
---
 .../agent-memory/frontend-developer/MEMORY.md |  1 +
 .../i18n-invisible-chars.md                   | 34 +++++++
 .../frontend-developer/qa-tripwire-tests.md   |  7 +-
 .../qa-integration-tester/MEMORY.md           |  2 +-
 .../pr-1959-inline-meta-content-loss.md       | 64 ++++++++++++-
 .claude/agent-memory/translator/MEMORY.md     | 10 ++-
 .../translator/abschlag-glossary-shortform.md | 33 +++++++
 .../translator/nbsp-inline-labels.md          | 45 ++++++++++
 .../reports/ReportContentEditor.module.css    | 16 +++-
 .../reports/ReportContentEditor.tsx           | 61 +++++++------
 client/src/i18n/de/budget.json                |  3 +-
 client/src/i18n/en/budget.json                |  3 +-
 client/src/i18n/glossary.json                 |  9 +-
 client/src/i18n/i18n.parity.test.ts           | 89 +++++++++++++++++++
 client/src/lib/reportPdf/overviewPdf.test.ts  | 43 +++++++++
 client/src/lib/reportPdf/overviewPdf.ts       | 20 +++++
 client/src/lib/reportPdf/realRender.test.ts   | 20 ++++-
 17 files changed, 424 insertions(+), 36 deletions(-)
 create mode 100644 .claude/agent-memory/frontend-developer/i18n-invisible-chars.md
 create mode 100644 .claude/agent-memory/translator/abschlag-glossary-shortform.md
 create mode 100644 .claude/agent-memory/translator/nbsp-inline-labels.md

diff --git a/.claude/agent-memory/frontend-developer/MEMORY.md b/.claude/agent-memory/frontend-developer/MEMORY.md
index aed9b8519..ecb0a2816 100644
--- a/.claude/agent-memory/frontend-developer/MEMORY.md
+++ b/.claude/agent-memory/frontend-developer/MEMORY.md
@@ -98,6 +98,7 @@ removing it will fail typecheck. Keep the export and note it as QA-deferred.
 - `budget-invoices.md` — open for budget/invoice/subsidy pages: BudgetSubNav, budget line APIs, invoice CRUD patterns, pipeline dashboard cards
 - `gantt-calendar.md` — open for Timeline/Gantt/Calendar/Milestone work: SVG architecture, drag hook (React 19 refs rule), arrow hover, lane allocation, constraints section, related E2E POM layout
 - `api-routing.md` — open when calling APIs or navigating: UserResponse/PaginatedResponse shapes, auth routes, dependency direction-swap semantics, navigation origin state
+- `i18n-invisible-chars.md` — open before adding an NBSP/invisible char to a locale file: literal-not-escape convention, why grep misses it, how to verify by render, and the identical-looking test failures it causes
 - `qa-tripwire-tests.md` — open when a task says a fix will flip an `it.failing` guard: why the flip signal can be swallowed by a shared helper, and what to do instead
 - `testing-mocks.md` — open when your change will break existing tests: ESM mock patterns, mock-block update checklist, Toast/context mocks, DefinePlugin globals in Jest
 - `git-tooling.md` — open on webpack config changes or git/CI trouble: worktree remotes, object corruption/permission recovery, npm audit policy, Prettier/CI format gap
diff --git a/.claude/agent-memory/frontend-developer/i18n-invisible-chars.md b/.claude/agent-memory/frontend-developer/i18n-invisible-chars.md
new file mode 100644
index 000000000..a1200de3d
--- /dev/null
+++ b/.claude/agent-memory/frontend-developer/i18n-invisible-chars.md
@@ -0,0 +1,34 @@
+---
+name: i18n-invisible-chars
+description: How to add and verify invisible characters (NBSP) in locale files — repo writes literal non-ASCII, and the resulting test/diff failures look like identical strings
+metadata:
+  type: project
+---
+
+Multi-word inline labels rendered into the PDF wrap at their internal space. A single-word label
+can't wrap; a two-word one breaks across lines and splits its own brackets (`(less` / `deposit)`).
+The fix is U+00A0 (NBSP) instead of the space — same glyph advance, so no geometry constant moves,
+and pdfmake's UAX-14 line breaker treats it as non-breaking.
+
+**Why the details matter:** the change is invisible in every textual channel, so it produces
+failures and diffs that look like no-ops and burn reviewer time.
+
+**How to apply:**
+
+- **Convention:** `client/src/i18n/**` uses **literal** non-ASCII characters (`…`, `—`, `·`, `↑`) —
+  there are zero `\uXXXX` escapes. Write a literal NBSP, not an escape. Insert it with a script
+  (`python3` + `u' '`), never by typing it, and verify with
+  `python3 -c "...print([hex(ord(c)) for c in value])"` — a hand-typed one silently lands as a plain
+  space.
+- **grep will not find it:** `grep -P '\xc2\xa0'` returns nothing in a UTF-8 locale (PCRE reads the
+  escapes as characters, not bytes). Use `LC_ALL=C grep -n $'\xc2\xa0' <file>`, or `git diff | cat -A`
+  (NBSP shows as `M-BM-`).
+- **`pdftotext` normalizes it back to a space**, so extraction can't confirm the character. Verify
+  the **break position** instead: render the real PDF, extract with `pdftotext -layout`, and check
+  the label lands on one line. Render a plain-space control alongside it — otherwise you can't tell
+  a fix from data that happened not to wrap.
+- **Warn about the test fallout explicitly.** Any test asserting the label literally now fails with
+  `Expected " (less deposit)" / Received " (less deposit)"` — visually identical. Always report the
+  codepoints and the exact replacement, or someone will read it as a flake.
+
+Related: [[qa-tripwire-tests]].
diff --git a/.claude/agent-memory/frontend-developer/qa-tripwire-tests.md b/.claude/agent-memory/frontend-developer/qa-tripwire-tests.md
index 14fff6f1d..631af7199 100644
--- a/.claude/agent-memory/frontend-developer/qa-tripwire-tests.md
+++ b/.claude/agent-memory/frontend-developer/qa-tripwire-tests.md
@@ -27,4 +27,9 @@ the exact helper line QA must update — do not "fix" the helper to make the fli
 (3) expect sibling tests in the same block to fail too, for the same reason (they assert the broken
 model at content-tree level, which is why they passed while the rendered PDF lost content).
 
-Related: [[../MEMORY.md]] "Refinement Workflow — QA Test Coordination".
+**Outcome on #1959 (closing the loop):** reporting the exact blocking helper line was the right call
+— QA re-derived the row count from the production packer itself, then converted the tripwire to a
+normal guard. Reporting "cannot flip, here is the line" got it resolved in one round; editing the
+helper myself would have put a developer edit in a QA-owned file for no gain.
+
+Related: [[../MEMORY.md]] "Refinement Workflow — QA Test Coordination", [[i18n-invisible-chars]].
diff --git a/.claude/agent-memory/qa-integration-tester/MEMORY.md b/.claude/agent-memory/qa-integration-tester/MEMORY.md
index 0a27ee35a..633f8b800 100644
--- a/.claude/agent-memory/qa-integration-tester/MEMORY.md
+++ b/.claude/agent-memory/qa-integration-tester/MEMORY.md
@@ -15,7 +15,7 @@
 
 ## Recent bug/story notes (2026-08)
 
-- [PR #1959 — inline meta content loss + `it.failing` tripwires](pr-1959-inline-meta-content-loss.md) (2026-08-03, RESOLVED) — prod defect found+fixed (unchunked meta in a `dontBreakRows` cell silently drops pages); **a tripwire is worthless if a shared helper bakes in the buggy assumption** — mine nearly stayed green through the fix; channel-equivalence is the threshold-free assertion; tree-level assertions cannot see this bug class; `grep` silently returns nothing on these test files (use `awk`).
+- [PR #1959 — inline meta content loss + `it.failing` tripwires](pr-1959-inline-meta-content-loss.md) (2026-08-03, RESOLVED) — prod defect found+fixed (unchunked meta in a `dontBreakRows` cell silently drops pages); **a tripwire is worthless if a shared helper bakes in the buggy assumption** — mine nearly stayed green through the fix; channel-equivalence is the threshold-free assertion; tree-level assertions cannot see this bug class; **write NBSP as `\u00A0` in test expectations, never a literal** (I smuggled one into the guard against it); keep literal+invariant at different levels; non-positive chunk budgets HANG not throw; `grep` silently returns nothing on these test files (use `awk`).
 - [Bug #1955 — echo-race harness + mutation probes](bug-1955-echo-race-harness.md) (2026-08-03) — echo must fire on the write's *resolve* (not the call) or the queue fix masks the guard and the test passes pre-fix; `rerender()` stands in for the optimistic `setPreferences`; 4 perl mutation probes prove each test guards a distinct part of the fix; never run repo-wide `npm run format` (38 unrelated files drift).
 
 - [Story #1930 — attachment tier rules](story-1930-attachment-tier.md) (2026-08-02) — new `attachmentTierUtils.test.ts` (100% cov, table+explicit-null pattern); sourceReportService.test.ts 16a/b/c/e → AC1/AC2/AC3/AC5; cross-report-type test needs a fresh invoice per report type; round 2 added a genuinely discriminating proof-of-funds fixture (deposit-only invoice, tagged+claimed deposit) after PO flagged the original AC3/16 pof blocks weren't change-detecting — inline-swap isolation technique to prove it fails on old logic.
diff --git a/.claude/agent-memory/qa-integration-tester/pr-1959-inline-meta-content-loss.md b/.claude/agent-memory/qa-integration-tester/pr-1959-inline-meta-content-loss.md
index 4f76d8f57..012efd02d 100644
--- a/.claude/agent-memory/qa-integration-tester/pr-1959-inline-meta-content-loss.md
+++ b/.claude/agent-memory/qa-integration-tester/pr-1959-inline-meta-content-loss.md
@@ -1,6 +1,6 @@
 ---
 name: pr-1959-inline-meta-content-loss
-description: PR #1959 (report PDF UX) reintroduced and then fixed the #1929 round-4 PDF content-loss regression; records the packUsageCellRows bound, the measured page-count saturation table, the it.failing tripwire technique AND the trap that nearly disarmed it (a helper that bakes in the buggy assumption), and the run-index gotcha for grey meta runs
+description: PR #1959 (report PDF UX) reintroduced and then fixed the #1929 round-4 PDF content-loss regression; records the packUsageCellRows bound, the page-count saturation table, the it.failing tripwire trap (a helper baking in the buggy assumption), the NBSP-must-be-escaped-in-test-expectations rule, literal-vs-invariant test layering, and the non-positive-budget hang
 metadata:
   type: project
 ---
@@ -92,6 +92,68 @@ Related: tree-level assertions **cannot see this bug class at all** — the drop
 the pdfmake content tree while the reader receives a truncated PDF. Where the claim is about what the
 reader receives, assert against rendered output (page count).
 
+## NBSP in inline PDF labels — and why a test expectation must escape it
+
+`ux-designer` found `depositReducedInlineLabel` was the only inline label in either locale with an
+internal space, so at 8pt in the 75pt allocated-amount column it wrapped at that space and split the
+brackets across lines (`€4,000.00 (less` / `deposit)` EN; `(Teilbetrag) (abzgl.` / `Abschlag)` DE) —
+in a document sent to a bank. Both locales now use U+00A0; same glyph advance, so no width or
+geometry constant moved.
+
+**In test expectations, write `\u00A0` as an escape, never paste the literal.** A literal NBSP and a
+literal `' '` are visually identical, so a failure reads as a flake, and the "obvious fix" — retyping
+a plain space — silently restores the wrap **with the unit suite green**, because the locale file and
+the expectation would agree with each other again. The escape is the only thing that makes the diff
+reviewable. (The repo convention of literal non-ASCII with zero `\uXXXX` escapes is about
+`client/src/i18n/**` *JSON*; it does not extend to test expectations, where visibility wins.)
+
+I violated this while writing the guard against it: the invariant test's own
+`c !== '<literal NBSP>'` filter went in with a literal NBSP instead of `'\u00A0'`. It passed, and was invisible.
+**Always scan a file's codepoints after writing NBSP-related assertions** —
+`node -e '[...fs.readFileSync(f,"utf8")].filter(c=>c.codePointAt(0)===0xA0)'`.
+
+### Literal vs invariant: keep both, at different levels
+
+Asked whether to replace the brittle literal with "contains no U+0020", the answer was both, because
+they catch disjoint things:
+
+- The **literal** in `realRender.test.ts` pins WIRING — that the deposit-reduced label of the *report*
+  language reaches the allocated-amount cell. It catches a cross-key/cross-locale mix-up (rendering
+  `splitNote` where `depositReducedNote` belongs). A "contains no space" check passes happily while
+  the wrong label renders.
+- The **invariant** in `client/src/i18n/i18n.parity.test.ts` pins TYPOGRAPHY, survives copy edits
+  (verified: rewording to `net of<NBSP>deposit` still passes), and covers every locale and every
+  future label automatically — the real regression surface.
+
+Scope it carefully: only keys that exist *solely* as bracketed inline labels
+(`splitInlineLabel`, `depositReducedInlineLabel`) get the hard no-U+0020 rule.
+`attachmentType.deposit` is **shared** — `buildReportContent.ts` also comma-joins it into flowing
+attachments-note prose, where NBSP would be over-reach — so it gets the weaker "is a single word"
+guard, which still fires if a translator makes it multi-word. The invariant rejects U+202F (narrow
+NBSP) too, deliberately: it is non-breaking but has a *narrower advance*, and "same advance, no
+geometry moved" is what made the fix safe to land unmeasured.
+
+## Non-positive chunk budgets hang rather than throw (now guarded)
+
+`splitIntoPageSafeChunks` with `maxChars <= 0` used to **spin forever**, not fail: the hard-split path
+does `token.slice(0, maxChars)`, and `slice(0, 0)` is `''`, so `rest` never shrinks (probe: 100,000
+iterations, `rest` unchanged). `packUsageCellRows` has the same shape — I confirmed it independently
+by replaying its loop bounded, and a mutation probe removing its guard **hung the jest runner**, which
+is itself the proof. Both now throw. When probing a guard whose absence is non-termination, replay the
+loop with a spin counter instead of running the suite, or the probe eats the timeout.
+
+Two subtle guard tests earned their place under mutation: an off-by-one (`<= 1`) is caught only by
+asserting the smallest *legal* budget (1) still works, and moving the guard after the
+`text.length <= maxChars` short-circuit is caught only by testing a non-positive budget with input
+short enough to short-circuit.
+
+## Mutation-probing JSON: target the path, not the string
+
+`s.replace('"deposit": "Deposit"', ...)` hit a *different* `deposit` key in another scope of the same
+file, so the probe reported a false negative on the guard. Mutate through
+`JSON.parse` → set the exact dotted path → `JSON.stringify`, and print the resulting value to confirm
+the mutation landed where intended.
+
 ## Gotcha: the grey meta run is never at a fixed run index
 
 `buildUsageTextRuns()` tokenizes prose into **one run per whitespace-delimited token**, so a cell's
diff --git a/.claude/agent-memory/translator/MEMORY.md b/.claude/agent-memory/translator/MEMORY.md
index 5c74d3d50..667997067 100644
--- a/.claude/agent-memory/translator/MEMORY.md
+++ b/.claude/agent-memory/translator/MEMORY.md
@@ -97,6 +97,11 @@ New `sourceReports.expand.*` (chevron-expand sub-tables for budget lines + depos
 - [Progress/status label style](progress-label-style.md) — two co-existing patterns for "X-ing… (Ns)" labels; pick the closer sibling key as precedent (`autoItemize.analyzing` → "Analysiere… ({{seconds}}s)")
 - [EMPTY_SELECTION backfill](empty-selection-error-code.md) — pre-existing error code with no locale entry in either en or de; verify via server code before assuming a key is "new"
 
+## PDF Inline-Label Constraints (#1917/#1959)
+
+- [NBSP for inline labels](nbsp-inline-labels.md) — multi-word inline PDF labels need U+00A0, not a regular space, or pdfmake wraps mid-bracket; includes the ad-hoc real-render/pdftoppm verification recipe
+- [Abschlag glossary short-form](abschlag-glossary-shortform.md) — PO-approved `Abschlag` short form recorded IN glossary.json (75pt column has no room for any qualifier); `split`'s 3 German forms deliberately un-pinned
+
 ## Audit Protocol History
 
 - [Audit pitfalls](audit-pitfalls.md) — incident history behind the mandatory 4-step full-coverage audit protocol: a parity-only audit missed 13 code-referenced keys (Area UI raw-key bug); loose substring greps flagged 52 false positives
@@ -106,8 +111,9 @@ New `sourceReports.expand.*` (chevron-expand sub-tables for budget lines + depos
 `sourceReports.editable.signatureLabel` → "Unterschrift"; `sourceReports.coverLetter.closing` → "Mit freundlichen Grüßen,"; `sourceReports.editable.closingLabel` → "Grußformel". Confirmed: neither "signature" nor "closing salutation" belongs in the glossary (grep across `glossary.json` for signature/closing/Gruß terms found nothing, and these are generic letter-writing vocabulary, not Cornerstone domain terms) — did not add.
 
 **Chrome-vs-content split, worked example**: `coverLetter.closing` and `editable.closingLabel` sit right next to each other in the UI but are translated by two different rules, and getting this backwards is the easy mistake:
-- `coverLetter.closing` = the letter's actual salutation text, rendered via `reportT()` into the PDF in the *report's* language → full letter-register translation ("Mit freundlichen Grüßen,").
-- `editable.closingLabel` = the read-only caption *above* that value in the step-5 editor, rendered via `t()` in the *interface's* language → UI chrome, translated like its sibling captions (`senderLabel`→"Absender", `subjectLabel`→"Betreff"), using the precise German business-letter term for "closing salutation" ("Grußformel", paired with "Anrede" for the opening salutation) rather than a literal "Schließen"/"Closing"-style calque.
+
+- `coverLetter.closing` = the letter's actual salutation text, rendered via `reportT()` into the PDF in the _report's_ language → full letter-register translation ("Mit freundlichen Grüßen,").
+- `editable.closingLabel` = the read-only caption _above_ that value in the step-5 editor, rendered via `t()` in the _interface's_ language → UI chrome, translated like its sibling captions (`senderLabel`→"Absender", `subjectLabel`→"Betreff"), using the precise German business-letter term for "closing salutation" ("Grußformel", paired with "Anrede" for the opening salutation) rather than a literal "Schließen"/"Closing"-style calque.
 - Rule of thumb: check the **call site**, not the JSON namespace path — both keys live under `sourceReports.*` but `t()` vs `reportT()` at the usage site is what actually decides chrome vs. content, per the established artifact-content-vs-edit-affordance convention from #1909/#1924.
 
 **Concurrent-worktree gotcha**: at spec time `closingLabel` didn't exist in `en/budget.json` yet (frontend-developer added it mid-session, confirmed via `git diff client/src/i18n/en/budget.json` showing the line appear between my two check-ins). When a spec says "check whether the en key already landed," re-check immediately before finalizing rather than trusting an earlier read — in a shared worktree the answer can flip within the same task.
diff --git a/.claude/agent-memory/translator/abschlag-glossary-shortform.md b/.claude/agent-memory/translator/abschlag-glossary-shortform.md
new file mode 100644
index 000000000..651b6a8f1
--- /dev/null
+++ b/.claude/agent-memory/translator/abschlag-glossary-shortform.md
@@ -0,0 +1,33 @@
+---
+name: abschlag-glossary-shortform
+description: "Abschlag" is a PO-approved short-form of the glossary term "Deposit" (Abschlagszahlung), permitted only under a measured space constraint — recorded in glossary.json itself
+metadata:
+  type: project
+---
+
+Issue #1917/#1959: `product-owner` ruled that `"Abschlag"` (short form) may be used instead of the
+glossary term `"Abschlagszahlung"` for the **deposit-reduced inline label**
+(`depositReducedInlineLabel`: "abzgl. Abschlag") in the report PDF's Allocated-amount table column.
+
+**Why there's no third option** (the PO's decisive point): the column budget is 75pt.
+`"Abschlagszahlung"` alone already consumes 72.85pt — so literally no qualifier of any length fits
+alongside it, not `abzgl.`, not `ohne`, not `./.`. Keeping the full term therefore forces dropping
+the qualifier entirely, which would silently collapse this label into the visually-similar but
+semantically-different **constituted**-deposit label (`(Abschlagszahlung)`, #1923 AC2.1). Those are
+different facts on a Verwendungsnachweis: "this row **is** a deposit" vs. "this amount is **reduced
+because** deposits were claimed separately" (context restored by #1965's revived footnote,
+`depositReducedFootnote` — do not delete that key).
+
+**Recorded directly in `client/src/i18n/glossary.json`** under the `Deposit` term's `de.shortForm` /
+`de.shortFormNote` fields (not just in this memory file) — the point is that a future glossary
+compliance sweep reading `glossary.json` alone must see the 75pt-column-budget reasoning and refuse
+to "fix" `abzgl. Abschlag` back to `abzgl. Abschlagszahlung` in good faith. See
+[[nbsp-inline-labels]] for the companion fix (NBSP inside the short-form label itself, to stop it
+wrapping mid-bracket) that this same PR round also required.
+
+**`split`'s three German forms — deliberately NOT glossary-pinned**: `anteilig` (adjective, e.g.
+"anteilig belassen") / `Anteil` (noun, "...zugeordneten Anteil") / `Teilbetrag` (noun,
+`splitInlineLabel`). The PO ruled the glossary exists to prevent _semantic_ divergence across
+translators/time, not surface-form variation — pinning one form here would force ungrammatical copy
+across an adjective and two nouns that all mean the same thing in context. Recorded on #1917 as
+"reject if proposed later" — do not re-propose a `Split`/`anteilig` glossary entry.
diff --git a/.claude/agent-memory/translator/nbsp-inline-labels.md b/.claude/agent-memory/translator/nbsp-inline-labels.md
new file mode 100644
index 000000000..e1805e1df
--- /dev/null
+++ b/.claude/agent-memory/translator/nbsp-inline-labels.md
@@ -0,0 +1,45 @@
+---
+name: nbsp-inline-labels
+description: Multi-word inline PDF labels must use U+00A0 (NBSP) instead of a regular space to prevent mid-bracket line wraps in pdfmake tables
+metadata:
+  type: project
+---
+
+Issue #1959 (PR review round): `depositReducedInlineLabel` (`de/budget.json`: "abzgl. Abschlag") was
+the only inline report-table label containing an internal space — every sibling label
+(`Teilbetrag`, `Abschlagszahlung`, `(Rückerstattung)`) is a single word. pdfmake wraps at any space
+by default, so this one broke across two lines with the opening/closing bracket separated —
+unacceptable in a document sent to a bank.
+
+**Fix applied**: replaced the regular space between "abzgl." and "Abschlag" with a literal U+00A0
+(non-breaking space, UTF-8 bytes `c2 a0`) in the JSON string — not a ` ` escape. Confirmed via
+`xxd` that the EN sibling key (`depositReducedInlineLabel`: "less deposit", owned by
+frontend-developer) already uses the same literal-NBSP convention in this repo, not the backslash
+escape form. NBSP has the identical glyph advance to a regular space, so this is a **zero-width-cost**
+fix — no page-geometry/column-width constant needs to change, and any prior width measurement in pt
+remains exactly valid.
+
+**Rule going forward**: any new multi-word inline label destined for a narrow pdfmake table column
+(report PDF overview/cover-letter tables) must use NBSP between its words, not a regular space,
+UNLESS the column is verified wide enough that the full label plus its preceding run (amount +
+label share one run stream) never approaches the column's declared pt width. Do not rely on a
+label-alone width calculation — pdfmake lays out the amount and the label as one continuous run, so
+the combined width is what determines wrapping, not the label's width in isolation (see
+[[abschlag-glossary-shortform]] for the width-budget details of this specific case).
+
+**Verification method**: width arithmetic alone missed this bug (see the same lesson at
+[[audit-pitfalls]] for the analogous "measured the wrong quantity" pattern in a different context).
+The only verification that caught it was rendering the _real_ pdfmake pipeline end-to-end and
+looking at the PDF. Reusable recipe (no committed harness exists for this — built ad hoc, deleted
+after use):
+
+1. Write a temporary `*.test.ts` file under `client/src/lib/reportPdf/` (Jest picks it up
+   automatically; delete it before handoff — it is not a real test, just a rendering harness).
+2. Reuse the real, unmocked pipeline exactly as `realRender.test.ts` does: real i18next instance
+   loaded with the actual `en/budget.json`/`de/budget.json`, real `buildReportContent` +
+   `generateReportPdf` (from `merge.ts`), real `loadPdfLibs()` (real pdfmake@0.3.11 + embedded
+   fonts). Needs the `Blob.prototype.arrayBuffer` FileReader polyfill from that file's `beforeAll`
+   (jsdom's Blob polyfill lacks `.arrayBuffer()`).
+3. `fs.writeFileSync('/tmp/x.pdf', Buffer.from(await result.blob.arrayBuffer()))`.
+4. `pdftoppm -png -r 200 /tmp/x.pdf /tmp/x` then `Read` the resulting PNG to visually inspect.
+5. Run with `NODE_OPTIONS=--experimental-vm-modules npx jest <path> --maxWorkers=1`.
diff --git a/client/src/components/reports/ReportContentEditor.module.css b/client/src/components/reports/ReportContentEditor.module.css
index ba129259f..2acfe7114 100644
--- a/client/src/components/reports/ReportContentEditor.module.css
+++ b/client/src/components/reports/ReportContentEditor.module.css
@@ -83,11 +83,25 @@
   flex-shrink: 0;
 }
 
+.columnToggleGroup {
+  display: flex;
+  flex-direction: column;
+  gap: var(--spacing-1);
+  margin-bottom: var(--spacing-3);
+}
+
+/* Preview-only disclaimer for the column toggles — these do not affect the generated PDF. */
+.columnToggleHint {
+  margin: 0;
+  font-size: var(--font-size-xs);
+  color: var(--color-text-muted);
+  font-style: italic;
+}
+
 .columnToggles {
   display: flex;
   flex-wrap: wrap;
   gap: var(--spacing-2);
-  margin-bottom: var(--spacing-3);
 }
 
 .columnToggle {
diff --git a/client/src/components/reports/ReportContentEditor.tsx b/client/src/components/reports/ReportContentEditor.tsx
index d019bc87f..5e3505d48 100644
--- a/client/src/components/reports/ReportContentEditor.tsx
+++ b/client/src/components/reports/ReportContentEditor.tsx
@@ -3,7 +3,7 @@
  * Handles field changes and resets via callbacks; no state management.
  */
 
-import { useState } from 'react';
+import { useId, useState } from 'react';
 import type { TFunction } from 'i18next';
 import type { InvoiceStatus } from '@cornerstone/shared';
 import type { ReportContent, ReportContentOverrides } from '../../lib/reportContent/index.js';
@@ -41,7 +41,12 @@ export function ReportContentEditor({
   // Helper: check if a field has been overridden
   const isFieldEdited = (key: string): boolean => key in overrides;
 
-  // Column visibility state
+  // Column visibility state. PREVIEW-ONLY: `hiddenColumns` is local to this component and is not
+  // exposed as a prop or callback — the generated PDF always contains every column. The hint
+  // rendered beside the toggles says so, because the control otherwise reads as "choose the
+  // report's columns" (every other control in this editor does change the PDF). Wiring these
+  // through to the PDF is a filed follow-up.
+  const columnHintId = useId();
   const [hiddenColumns, setHiddenColumns] = useState<Set<ColumnKey>>(new Set());
   const toggleColumn = (col: ColumnKey) => {
     setHiddenColumns((prev) => {
@@ -192,29 +197,35 @@ export function ReportContentEditor({
       {/* Report Table */}
       <div className={styles.tableHeadingRow}>
         <h3 className={styles.tableHeading}>{t('sourceReports.editable.tableHeading')}</h3>
-        <div
-          className={styles.columnToggles}
-          role="group"
-          aria-label={t('sourceReports.editable.columnVisibilityLabel')}
-        >
-          {(
-            [
-              ['vendor', content.labels.vendor],
-              ['invoiceNumber', content.labels.invoiceNumber],
-              ['date', content.labels.date],
-              ...(content.isOverview
-                ? [['status', content.labels.status] as [ColumnKey, string]]
-                : []),
-              ['invoiceAmount', content.labels.invoiceAmount],
-              ['allocatedAmount', content.labels.allocatedAmount],
-              ['usage', content.labels.usage],
-            ] as [ColumnKey, string][]
-          ).map(([col, label]) => (
-            <label key={col} className={styles.columnToggle}>
-              <input type="checkbox" checked={show(col)} onChange={() => toggleColumn(col)} />
-              {label}
-            </label>
-          ))}
+        <div className={styles.columnToggleGroup}>
+          <p id={columnHintId} className={styles.columnToggleHint}>
+            {t('sourceReports.editable.columnVisibilityHint')}
+          </p>
+          <div
+            className={styles.columnToggles}
+            role="group"
+            aria-label={t('sourceReports.editable.columnVisibilityLabel')}
+            aria-describedby={columnHintId}
+          >
+            {(
+              [
+                ['vendor', content.labels.vendor],
+                ['invoiceNumber', content.labels.invoiceNumber],
+                ['date', content.labels.date],
+                ...(content.isOverview
+                  ? [['status', content.labels.status] as [ColumnKey, string]]
+                  : []),
+                ['invoiceAmount', content.labels.invoiceAmount],
+                ['allocatedAmount', content.labels.allocatedAmount],
+                ['usage', content.labels.usage],
+              ] as [ColumnKey, string][]
+            ).map(([col, label]) => (
+              <label key={col} className={styles.columnToggle}>
+                <input type="checkbox" checked={show(col)} onChange={() => toggleColumn(col)} />
+                {label}
+              </label>
+            ))}
+          </div>
         </div>
       </div>
       <div className={styles.tableWrapper}>
diff --git a/client/src/i18n/de/budget.json b/client/src/i18n/de/budget.json
index 572cf2640..7d2445cbd 100644
--- a/client/src/i18n/de/budget.json
+++ b/client/src/i18n/de/budget.json
@@ -1216,6 +1216,7 @@
       "coverLetterHeading": "Anschreiben",
       "tableHeading": "Berichtstabelle",
       "columnVisibilityLabel": "Spalten ein-/ausblenden",
+      "columnVisibilityHint": "Nur Vorschau – das erzeugte PDF enthält immer alle Spalten",
       "enhanceWithAi": "Mit KI verbessern",
       "enhanceWithAiDescription": "Ersetzt die unten stehenden Verwendungstexte und das Anschreiben durch KI-generierte Inhalte. Vorgenommene Bearbeitungen gehen dabei verloren.",
       "generating": "Generiere… ({{seconds}}s)",
@@ -1259,7 +1260,7 @@
       "attachmentsNoteNoType_other": "{{count}} Anhänge",
       "depositReducedFootnote": "Diese Position berücksichtigt separat eingereichte Abschlagszahlungen.",
       "splitInlineLabel": "Teilbetrag",
-      "depositReducedInlineLabel": "abzgl. Abschlag"
+      "depositReducedInlineLabel": "abzgl. Abschlag"
     },
     "sourceType": {
       "bank_loan": "Bankdarlehen",
diff --git a/client/src/i18n/en/budget.json b/client/src/i18n/en/budget.json
index 1921c8803..67061da91 100644
--- a/client/src/i18n/en/budget.json
+++ b/client/src/i18n/en/budget.json
@@ -1216,6 +1216,7 @@
       "coverLetterHeading": "Cover Letter",
       "tableHeading": "Report Table",
       "columnVisibilityLabel": "Show/hide columns",
+      "columnVisibilityHint": "Preview only — the generated PDF always includes every column",
       "enhanceWithAi": "Enhance with AI",
       "enhanceWithAiDescription": "Replaces the usage descriptions and cover letter below with AI-generated content. Any edits you've made will be discarded.",
       "generating": "Generating… ({{seconds}}s)",
@@ -1259,7 +1260,7 @@
       "splitFootnote": "Amount shown reflects only the portion allocated to this source.",
       "depositReducedFootnote": "This position reflects deposits claimed separately.",
       "splitInlineLabel": "partial",
-      "depositReducedInlineLabel": "less deposit"
+      "depositReducedInlineLabel": "less deposit"
     },
     "sourceType": {
       "bank_loan": "Bank Loan",
diff --git a/client/src/i18n/glossary.json b/client/src/i18n/glossary.json
index 760896f8a..5523efec7 100644
--- a/client/src/i18n/glossary.json
+++ b/client/src/i18n/glossary.json
@@ -24,7 +24,14 @@
     "Area": { "de": { "singular": "Bereich", "plural": "Bereiche" } },
     "Trade": { "de": { "singular": "Gewerk", "plural": "Gewerke" } },
     "Unassigned": { "de": { "singular": "Nicht zugewiesen" } },
-    "Deposit": { "de": { "singular": "Abschlagszahlung", "plural": "Abschlagszahlungen" } },
+    "Deposit": {
+      "de": {
+        "singular": "Abschlagszahlung",
+        "plural": "Abschlagszahlungen",
+        "shortForm": "Abschlag",
+        "shortFormNote": "product-owner-approved (#1917) short-form variant of 'Abschlagszahlung', permitted ONLY under a measured space constraint — do not use elsewhere. Origin: the report PDF's 'abzgl. Abschlag' inline label lives in a 75pt table column. 'Abschlagszahlung' alone already measures 72.85pt, so no qualifier of any length fits alongside it — not 'abzgl.', not 'ohne', not './.'. There is no third option: keeping the full term forces dropping the qualifier, which collapses this label into the separate constituted-deposit label '(Abschlagszahlung)' (#1923 AC2.1) — a different fact ('this row IS a deposit' vs. 'this amount is reduced because deposits were claimed separately'). A future compliance sweep must NOT revert this to 'Abschlagszahlung' without re-measuring the 75pt column budget first; doing so silently breaks the PDF (label wraps mid-bracket, #1959)."
+      }
+    },
     "Refund": { "de": { "singular": "Rückerstattung", "plural": "Rückerstattungen" } },
     "Final payment": { "de": { "singular": "Schlusszahlung" } },
     "Draft": { "de": { "singular": "Entwurf", "plural": "Entwürfe" } },
diff --git a/client/src/i18n/i18n.parity.test.ts b/client/src/i18n/i18n.parity.test.ts
index 189b1ca52..2d943718e 100644
--- a/client/src/i18n/i18n.parity.test.ts
+++ b/client/src/i18n/i18n.parity.test.ts
@@ -221,3 +221,92 @@ describe('i18n duplicate JSON key guard', () => {
     }
   }
 });
+
+// ─── #1959: inline PDF labels must not be able to wrap mid-label ───────────
+//
+// `overviewPdf.ts` renders three labels as a bracketed inline suffix on the report's
+// allocated-amount cell — `(${label})` at 8pt in a 75pt-wide column. If such a label contains an
+// ordinary space, pdfmake wraps at it and splits the brackets across two lines: `ux-designer`
+// measured a real render producing `€4,000.00 (less` / `deposit)` in EN and `(Teilbetrag) (abzgl.` /
+// `Abschlag)` in DE, in a document sent to a bank. The fix was U+00A0 (same glyph advance, so no
+// width or geometry constant moved).
+//
+// WHY THIS LIVES HERE, and not only as a literal in realRender.test.ts. Both exist, deliberately,
+// because they catch different things:
+//   - The literal in realRender.test.ts pins WIRING — that the deposit-reduced label of the REPORT
+//     language reaches the allocated-amount cell. It is what catches a cross-key or cross-locale
+//     mix-up (rendering `splitNote` where `depositReducedNote` belongs). A "contains no space"
+//     assertion there would happily pass while the wrong label rendered.
+//   - This test pins the TYPOGRAPHIC INVARIANT, and it is the one that survives copy edits: reword
+//     the labels freely and it still holds. It also covers every locale and every future label
+//     automatically, which is the real regression surface — the next translator to add a
+//     two-word inline label, or anyone who "fixes" an NBSP mismatch by retyping a plain space
+//     (the two are visually identical), fails here.
+describe('#1959 inline PDF label typography', () => {
+  const budget = NAMESPACES.find((ns) => ns.name === 'budget')!;
+  // Iterate the namespace entry's own locale keys so a newly added locale is covered without
+  // touching this test (the registry above is the single source of truth for which locales exist).
+  const localeBundles = Object.entries(budget).filter(([k]) => k !== 'name') as [
+    string,
+    Record<string, unknown>,
+  ][];
+
+  /** Reads a dot-delimited string leaf out of a locale bundle, failing loudly if it is absent. */
+  function label(bundle: Record<string, unknown>, dottedPath: string): string {
+    let node: unknown = bundle;
+    for (const segment of dottedPath.split('.')) {
+      if (node === null || typeof node !== 'object') {
+        throw new Error(`${dottedPath} — segment "${segment}" has no parent object`);
+      }
+      node = (node as Record<string, unknown>)[segment];
+    }
+    if (typeof node !== 'string') {
+      throw new Error(`${dottedPath} is missing or not a string (got ${typeof node})`);
+    }
+    return node;
+  }
+
+  it('covers every registered locale (guards against this suite silently testing only one)', () => {
+    expect(localeBundles.map(([locale]) => locale).sort()).toEqual(['de', 'en']);
+  });
+
+  // These two keys exist for no purpose other than the bracketed inline suffix, so the constraint
+  // is unconditional for them.
+  for (const key of ['splitInlineLabel', 'depositReducedInlineLabel']) {
+    for (const [locale, bundle] of localeBundles) {
+      it(`${locale}: sourceReports.table.${key} contains no breaking space (U+0020) — it would split the brackets across lines`, () => {
+        const value = label(bundle, `sourceReports.table.${key}`);
+        // Positive anchor first: a non-empty label, so the absence assertion below cannot pass on
+        // an empty or missing string.
+        expect(value.length).toBeGreaterThan(0);
+        expect(value).not.toContain(' ');
+        // Any whitespace this label does contain must be exactly U+00A0. Asserted on the whole
+        // string rather than only U+0020, so a "fix" swapping in some other whitespace is caught:
+        // a tab or U+2009 thin space would still break the line, and even U+202F (narrow NBSP,
+        // which does not break) is rejected deliberately — it has a NARROWER glyph advance than
+        // U+00A0, and "same advance, so no width or geometry constant moved" is precisely what
+        // made this fix safe to land without re-measuring the column budgets.
+        // '\u00A0' as an escape, not a literal — this file is the one telling everyone else
+        // that an invisible NBSP is the hazard, so it must not smuggle one into its own
+        // source. (It did, on first writing. Caught by scanning this file's own codepoints.)
+        const breaking = [...value].filter((c) => /\s/u.test(c) && c !== '\u00A0');
+        expect(breaking).toEqual([]);
+      });
+    }
+  }
+
+  // `attachmentType.deposit` is the third label rendered bracketed inline — but unlike the two
+  // above it is SHARED: buildReportContent.ts also comma-joins it into the flowing attachments-note
+  // prose ("2 attachments: Quotation, Invoice"), where a non-breaking space would be over-reach.
+  // So it is held to the weaker invariant that actually matters: it is a single word today, hence
+  // has nothing to wrap at. If this ever fails, that is a decision point, not a typo — the new
+  // multi-word value needs the same U+00A0 treatment as depositReducedInlineLabel, because it
+  // renders in the same narrow bracketed slot.
+  for (const [locale, bundle] of localeBundles) {
+    it(`${locale}: sourceReports.table.attachmentType.deposit is a single word, so the bracketed inline form cannot wrap`, () => {
+      const value = label(bundle, 'sourceReports.table.attachmentType.deposit');
+      expect(value.length).toBeGreaterThan(0);
+      expect(value).not.toMatch(/\s/u);
+    });
+  }
+});
diff --git a/client/src/lib/reportPdf/overviewPdf.test.ts b/client/src/lib/reportPdf/overviewPdf.test.ts
index e05ed69ac..82ecc0927 100644
--- a/client/src/lib/reportPdf/overviewPdf.test.ts
+++ b/client/src/lib/reportPdf/overviewPdf.test.ts
@@ -355,6 +355,34 @@ describe('splitIntoPageSafeChunks (scenario 8)', () => {
     expect(splitIntoPageSafeChunks(exact, 50)).toEqual([exact]);
   });
 
+  // #1959 fix round. A non-positive budget used to HANG rather than fail: the hard-split path does
+  // `token.slice(0, maxChars)`, and `slice(0, 0)` is `''`, so `rest` never shrinks and the loop
+  // spins forever (a probe hit 100,000 iterations before being killed). A clamp was rejected in
+  // favour of throwing, because a silently clamped budget is a bound nobody can see. These tests
+  // also serve as termination proofs: if the guard regressed to a clamp-or-spin, they would time
+  // out rather than fail fast.
+  it.each([0, -1, -650])(
+    'throws on a non-positive maxChars (%i) instead of looping forever, and names the offending value',
+    (maxChars) => {
+      expect(() => splitIntoPageSafeChunks('some text that exceeds any budget', maxChars)).toThrow(
+        `splitIntoPageSafeChunks: maxChars must be positive, got ${maxChars}`,
+      );
+    },
+  );
+
+  it('rejects a non-positive maxChars even when the text would short-circuit as already-fitting', () => {
+    // `text.length <= maxChars` returns early, so an unguarded implementation would accept a
+    // nonsense budget for short input and only blow up later on long input — the guard must come
+    // first so the contract is the same for every input.
+    expect(() => splitIntoPageSafeChunks('', 0)).toThrow(/maxChars must be positive/);
+    expect(() => splitIntoPageSafeChunks('x', -5)).toThrow(/maxChars must be positive/);
+  });
+
+  it('accepts the smallest legal budget (1) and still terminates, one character per chunk', () => {
+    // Boundary on the legal side of the guard: 1 is the tightest budget that can make progress.
+    expect(splitIntoPageSafeChunks('abc', 1)).toEqual(['a', 'b', 'c']);
+  });
+
   it('(b) input with clean word boundaries splits into <= maxChars chunks that rejoin (chunks.join("")) to the exact original', () => {
     const text = Array.from({ length: 40 }, (_, i) => `word${i}`).join(' ');
     const maxChars = 30;
@@ -447,6 +475,21 @@ describe('packUsageCellRows (#1959 fix round: one page-safe budget for the whole
     return rows.flatMap((row, i) => (row.some((s) => s.meta) ? [i] : []));
   }
 
+  // #1959 fix round: same non-positive-budget contract as splitIntoPageSafeChunks (which this
+  // delegates to for hard splits, and which would otherwise spin). Throws rather than clamps.
+  it.each([0, -1, -650])(
+    'throws on a non-positive maxChars (%i) rather than looping or clamping, and names the offending value',
+    (maxChars) => {
+      expect(() =>
+        packUsageCellRows([{ text: 'prose' }, { text: '\nmeta', meta: true }], maxChars),
+      ).toThrow(`packUsageCellRows: maxChars must be positive, got ${maxChars}`);
+    },
+  );
+
+  it('rejects a non-positive maxChars even for an empty stream, which would otherwise short-circuit', () => {
+    expect(() => packUsageCellRows([], 0)).toThrow(/maxChars must be positive/);
+  });
+
   it('returns a single row, segments unchanged, when the whole stream fits the budget (the dominant case must not grow rows)', () => {
     const segments: UsageCellSegment[] = [
       { text: 'Kitchen' },
diff --git a/client/src/lib/reportPdf/overviewPdf.ts b/client/src/lib/reportPdf/overviewPdf.ts
index 0732730f6..c4999357b 100644
--- a/client/src/lib/reportPdf/overviewPdf.ts
+++ b/client/src/lib/reportPdf/overviewPdf.ts
@@ -176,8 +176,17 @@ export const MAX_SAFE_USAGE_CHUNK_CHARS = 650;
  * except when a single whitespace-free run itself exceeds maxChars, in which case it is
  * hard-split so the algorithm always makes forward progress. Joining the returned chunks with
  * '' reconstructs `text` exactly — no character, including whitespace, is ever dropped (#1929 I1).
+ *
+ * Throws on a non-positive `maxChars`. Without the guard the hard-split loop spins forever
+ * (`token.slice(0, 0)` is `''` and `token.slice(0)` is `token`, so it never makes progress) — it
+ * does NOT fail loudly on its own. Unreachable today (every caller passes a module constant), but
+ * that stops being true the moment a budget is derived from column visibility, so it fails fast
+ * rather than clamping: a silently clamped budget is a bound nobody can see.
  */
 export function splitIntoPageSafeChunks(text: string, maxChars: number): string[] {
+  if (maxChars <= 0) {
+    throw new Error(`splitIntoPageSafeChunks: maxChars must be positive, got ${maxChars}`);
+  }
   if (text.length <= maxChars) return [text];
 
   const tokens = text.split(/(\s+)/); // capturing group keeps whitespace as its own tokens
@@ -240,11 +249,22 @@ export interface UsageCellSegment {
  * Lossless (#1929 I1): concatenating every returned segment's `text`, in order, reconstructs the
  * input stream exactly — no character, including whitespace, is added or dropped. Empty segments
  * are preserved so an empty `usageText` still renders its own (empty) leading run.
+ *
+ * Throws on a non-positive `maxChars`, which is not a survivable budget: with no room for even one
+ * character the packing loop cannot make progress (`remaining <= 0` flushes a row that is already
+ * empty, then retries the same state forever) — note this hangs in the packer's OWN loop, before it
+ * ever delegates to splitIntoPageSafeChunks, so both functions need their own guard. Unreachable
+ * while the budget is a module constant, but MAX_SAFE_USAGE_CHUNK_CHARS was measured against the
+ * 7-column geometry, so any future per-shape or column-visibility-derived budget must fail loudly
+ * here rather than be clamped into a bound nobody can see.
  */
 export function packUsageCellRows(
   segments: UsageCellSegment[],
   maxChars: number,
 ): UsageCellSegment[][] {
+  if (maxChars <= 0) {
+    throw new Error(`packUsageCellRows: maxChars must be positive, got ${maxChars}`);
+  }
   const rows: UsageCellSegment[][] = [];
   let current: UsageCellSegment[] = [];
   let used = 0;
diff --git a/client/src/lib/reportPdf/realRender.test.ts b/client/src/lib/reportPdf/realRender.test.ts
index b340f9dc5..ac155f940 100644
--- a/client/src/lib/reportPdf/realRender.test.ts
+++ b/client/src/lib/reportPdf/realRender.test.ts
@@ -959,7 +959,21 @@ describe('report PDF pipeline — real, unmocked end-to-end render', () => {
           formattersFor('en-US'),
           {
             depositLabel: ' (Deposit)',
-            depositReducedLabel: ' (less deposit)',
+            // NON-BREAKING SPACE, written as an escape ON PURPOSE. `depositReducedInlineLabel` is
+            // the only inline label in either locale with an internal space, and at 8pt in the
+            // narrow allocated-amount column it wrapped at that space — putting the opening and
+            // closing bracket on different lines ("€4,000.00 (less" / "deposit)") in a document
+            // sent to a bank. Both locales now use U+00A0, which has the same glyph advance, so no
+            // width or geometry constant moved.
+            //
+            // Do NOT "fix" a mismatch here by retyping a plain space: the literal NBSP and a
+            // literal ' ' are visually identical, so a diff looks clean while the wrap silently
+            // returns — and the unit suite stays GREEN, because the locale file and this
+            // expectation would agree with each other again. The escape is what makes the
+            // difference reviewable. See also the locale-level invariant in
+            // client/src/i18n/i18n.parity.test.ts, which fails if any inline label regains a
+            // breaking space.
+            depositReducedLabel: ' (less\u00A0deposit)',
             splitLabel: ' (partial)',
             // The footnote WORDINGS that must no longer appear anywhere in the tree.
             goneFootnotes: [
@@ -973,7 +987,9 @@ describe('report PDF pipeline — real, unmocked end-to-end render', () => {
           formattersFor('de-DE'),
           {
             depositLabel: ' (Abschlagszahlung)',
-            depositReducedLabel: ' (abzgl. Abschlag)',
+            // NON-BREAKING SPACE — see the en block above for why this is an escape, not a literal.
+            // DE wrapped as "(Teilbetrag) (abzgl." / "Abschlag)" before the fix.
+            depositReducedLabel: ' (abzgl.\u00A0Abschlag)',
             splitLabel: ' (Teilbetrag)',
             goneFootnotes: [
               'Diese Position berücksichtigt separat eingereichte Abschlagszahlungen.',