diff --git a/.claude/agent-memory/e2e-test-engineer/MEMORY.md b/.claude/agent-memory/e2e-test-engineer/MEMORY.md
index acde9174d..dcd665f33 100644
--- a/.claude/agent-memory/e2e-test-engineer/MEMORY.md
+++ b/.claude/agent-memory/e2e-test-engineer/MEMORY.md
@@ -6,7 +6,8 @@
 
 - [general-e2e-patterns.md](general-e2e-patterns.md) — cross-cutting waits/timing, viewport timeouts, strict-mode anti-patterns, DataTable migration fallout, Gantt touch, breadcrumbs, print, dashboard cards, headless-shell PDF-iframe limitation → CSP header+console verification pattern (blob-fetch leg removed, connect-src pitfall), key file locations.
 - [e2e-pom-patterns.md](e2e-pom-patterns.md) — Page Object Model conventions and stable-locator strategies.
-- [e2e-parallel-isolation.md](e2e-parallel-isolation.md) — `testPrefix` fixture, parallel-safe data isolation, serial-mode tests.
+- [e2e-parallel-isolation.md](e2e-parallel-isolation.md) — `testPrefix` fixture, parallel-safe data isolation, serial-mode tests (serial is **superseded** for preference rows — see next line).
+- [isolated-user-fixture.md](isolated-user-fixture.md) — `e2e/fixtures/isolatedUser.ts` dedicated-user opt-in for preference-mutating specs (#1957); **`test.use()` reshuffles shard membership suite-wide → unrelated shards go red**; login rate limit 20/15min, soft user delete, worker-scoped-option-in-describe ban, `--list` as the only local validation.
 - [known-flakes-and-regressions.md](known-flakes-and-regressions.md) — triaged log of flaky tests, pre-existing CI failures, and production regressions caught by E2E. **Check this before re-triaging a failure.**
 - [flake-patterns.md](flake-patterns.md) — flake-avoidance patterns moved from the implementation checklist: Konva canvas coordinates, `test.slow()` timeout negation, locale timing after reload, shard redistribution, stale cache-warmup CI. Check when writing timing-sensitive tests or adding spec files.
 - [auto-itemize-and-invoices-e2e.md](auto-itemize-and-invoices-e2e.md) — AutoItemizePage, PaperlessInvoiceReviewPage, invoice budget lines, Paperless mocking, vendor reassignment.
diff --git a/.claude/agent-memory/e2e-test-engineer/e2e-parallel-isolation.md b/.claude/agent-memory/e2e-test-engineer/e2e-parallel-isolation.md
index caa1a4903..7b38a2093 100644
--- a/.claude/agent-memory/e2e-test-engineer/e2e-parallel-isolation.md
+++ b/.claude/agent-memory/e2e-test-engineer/e2e-parallel-isolation.md
@@ -94,6 +94,13 @@ expect(countAfter).toBe(countBefore); // unchanged (not toBe(DEFAULT_CATEGORIES.
 
 ## Shared State Tests: Serial Mode
 
+> **Superseded for per-user _preference_ rows (Issue #1957).** Serial mode only orders a file
+> against itself, so it cannot protect `locale` / `dashboard.hiddenCards` / `table.*.columns`
+> from writes by another file in another worker. Those specs now use a dedicated user via
+> `e2e/fixtures/isolatedUser.ts` — see [isolated-user-fixture.md](isolated-user-fixture.md).
+> Serial mode is still the right tool for shared-admin _identity_ mutations (display name,
+> password, role) as listed below.
+
 Tests that modify the shared admin user (display name, password, role) cannot use `testPrefix`
 since there's only one admin user. Use `test.describe.configure({ mode: 'serial' })` inside the describe block:
 
diff --git a/.claude/agent-memory/e2e-test-engineer/general-e2e-patterns.md b/.claude/agent-memory/e2e-test-engineer/general-e2e-patterns.md
index daaf718e2..53aad8699 100644
--- a/.claude/agent-memory/e2e-test-engineer/general-e2e-patterns.md
+++ b/.claude/agent-memory/e2e-test-engineer/general-e2e-patterns.md
@@ -153,6 +153,33 @@ technique that touches network/fetch is a safe workaround — check which specif
 (`connect-src` vs `frame-src` vs `script-src`, etc.) governs the technique you're about to add,
 since they're independently configured and a fix for one doesn't imply permission under another.**
 
+## Never assert a suite-global DB precondition — scope list assertions to your own data (#1957, 2026-08-02)
+
+An assertion like "the `?areaId=__none__` list is empty" is really "no area-less household item
+exists anywhere in the shared DB" — and 48 of the suite's 62 `createHouseholdItemViaApi()` calls,
+across 12 spec files, pass no areaId. Under `fullyParallel: true` such a test passes only while
+no such spec shares its shard — pure shard luck, and it breaks the moment shard boundaries move
+(which any `test.use()` addition anywhere in the suite does; see
+[isolated-user-fixture.md](isolated-user-fixture.md)). It reads as a mysterious, unrelated
+failure.
+
+**Pattern:** AND the filter under test with a search scoped to the test's own entity names —
+e.g. `?areaId=__none__&q=<testPrefix + scenario token>` on any DataTable page (`useTableState`
+hydrates `q` from the URL, and the list services AND it with the other filters). This keeps the
+empty-state expectations valid, because `q` counts towards DataTable's `hasActiveFilters`
+(→ "No items match the current filters" + Clear Filters button). URL-driven, so it works on
+mobile where filter popovers are CSS-hidden.
+
+**Always pair it with a positive control**, or the fix creates a vacuous assertion: navigate
+with `q` alone first and assert the seeded item IS listed. Otherwise an empty result is
+indistinguishable from "the search matched nothing" and the test passes even if the filter
+under test does nothing. Worked example: Scenario 4 of
+`e2e/tests/household-items/no-area-filter.spec.ts`.
+
+Sibling assertions in the same family are fine when they are per-entity rather than global —
+`area-filter.spec.ts` Scenario 6 filters by a freshly created area ID it owns, so no other spec
+can put an item in it. Check which shape an assertion has before "fixing" it.
+
 ## Key file locations
 
 - Test fixtures: `e2e/fixtures/auth.ts` (testPrefix, authenticatedPage)
diff --git a/.claude/agent-memory/e2e-test-engineer/isolated-user-fixture.md b/.claude/agent-memory/e2e-test-engineer/isolated-user-fixture.md
new file mode 100644
index 000000000..a266de578
--- /dev/null
+++ b/.claude/agent-memory/e2e-test-engineer/isolated-user-fixture.md
@@ -0,0 +1,85 @@
+---
+name: isolated-user-fixture
+description: e2e/fixtures/isolatedUser.ts — how to give a preference-mutating spec its own user; why opting a file in reshuffles shard membership suite-wide and reddens unrelated shards; plus the login rate limit and soft-delete facts that dictate per-worker vs per-test scope
+metadata:
+  type: project
+---
+
+`e2e/fixtures/isolatedUser.ts` (added for Issue #1957) is the canonical way to stop a spec
+from reading/writing per-user preference rows on the shared admin
+(`test-results/.auth/admin.json`). Import `test`/`expect` from it and opt in:
+
+- `test.use({ isolatedUserPerWorker: { emailPrefix: 'dash' } })` — **file scope only**, one
+  dedicated user per worker. Preferred.
+- `test.use({ isolatedUserPerTest: { emailPrefix: 'x' } })` — file _or_ describe scope, fresh
+  user per test. Use when only a few tests inside a shared-admin file need isolation.
+
+It works by overriding the built-in **`storageState` option**, so plain `page`/`context` are
+already the dedicated user (no `scopedPage` rewrites) _and_ Playwright keeps trace/video/
+screenshot artifacts — a hand-made `browser.newContext()` silently loses those. An `auto`
+guard fixture checks `/api/auth/me` once per test so a silent regression of the override
+fails loudly instead of quietly reverting to shared-admin writes.
+
+**Why:** `LocaleContext.syncWithServer` treats the server as authoritative — it applies the
+server locale _and_ deletes the victim's `localStorage.locale`, so a concurrent PATCH from
+another file flips a running test's language permanently. Same shape for
+`dashboard.hiddenCards` and `table.*.columns`. `mode: 'serial'` cannot fix it (orders a file
+against itself only).
+
+**How to apply — constraints that decide the scope, verified against the server:**
+
+- `POST /api/auth/login` is rate-limited **20 requests / 15 min per IP**
+  (`server/src/routes/auth.ts`); every worker in a shard shares the bucket (`global: false`,
+  so only explicitly-configured routes are limited). Per-worker = 1 login per worker;
+  per-test multiplies fast. Budget it before converting a big file.
+- `DELETE /api/users/:id` is a **soft delete** (`deactivateUser` + `destroyUserSessions`) —
+  the `user_preferences` row survives, so the issue's "ON DELETE CASCADE cleans up" claim is
+  wrong. Isolation still holds (the account can never be logged into again), but users
+  accumulate.
+- User accumulation is bounded **per shard, not per run**: each shard is its own CI job with
+  its own container and DB (`containers/setup.ts` runs in globalSetup). So the 100-row
+  `/settings/users` page (`sortBy: null`, insertion order, scanned row-by-row by
+  `edit-user.spec.ts` / `deactivate-user.spec.ts`) is never remotely in reach — my original
+  worry was overstated. The case for per-worker rests on the login rate limit above.
+- Keep the string "admin" out of generated e-mails/display names —
+  `search-users.spec.ts` asserts every row matching a search for "Admin" has "admin" in its
+  name column.
+- Admin-gating is narrow: server `requireRole('admin')` only on `/api/users` mutations +
+  `/api/backups/*`; client only the Settings sub-nav "User Management"/"Backups" tabs and
+  work-item note edit/delete. So `role: 'member'` (the default) is fine for almost everything.
+
+**Playwright mechanics learned here (all verified in `node_modules/playwright/lib`):**
+
+- A **worker-scoped option cannot be set from inside a `describe`** — Playwright throws
+  "Cannot use({ x }) in a describe group, because it forces a new worker." That is the only
+  reason `isolatedUserPerTest` exists.
+- Option values participate in the worker hash, so each distinct `test.use()` value forces
+  its own worker group (extra worker restarts, a few seconds per shard).
+- **`test.use()` reshuffles shard MEMBERSHIP across the whole suite — expect an unrelated
+  shard to go red.** `createTestGroups` (`node_modules/playwright/lib/runner/index.js:2251`)
+  buckets tests by `test._workerHash` **first** and emits groups in bucket insertion order;
+  `filterForShard` (`:2321`) then walks that list and slices by cumulative test count. So
+  changing one file's worker hash moves **every** shard boundary in the suite, even with the
+  total test count unchanged. Any latent cross-file shared-state assumption in a newly
+  co-located pair then fails, looking completely unrelated to the change. This is what bit
+  #1957: `no-area-filter.spec.ts` moved from shard 15 to shard 14 next to
+  `area-filter.spec.ts`, whose Scenarios 3/5/6 hold area-less household items, and its
+  suite-global "no unassigned household item exists" precondition became unreachable.
+  **Diagnostic before blaming the change:** diff shard membership between base and head with
+  `npx playwright test --list --shard=N/16` — extract the base tree via
+  `git archive <base-sha> e2e | tar -x -C /tmp/base` and symlink `node_modules` into it.
+  Never respond by pinning or reordering shard assignment; fix the test that assumes
+  suite-global state (scope it to its own data — see
+  [general-e2e-patterns.md](general-e2e-patterns.md) and the Scenario 4 comment in
+  `e2e/tests/household-items/no-area-filter.spec.ts` for the `&q=`+`testPrefix` pattern).
+- `_combinedContextOptions` depends on the `storageState` _fixture_, so overriding that option
+  really does re-point `page`/`context` (not a silent no-op).
+- Project context options (incl. `baseURL`) are merged into `browser.newContext()` /
+  `request.newContext()` only by instrumentation installed by the test-scoped
+  `_setupArtifacts` fixture. A **worker-scoped** fixture runs before that, so it must pass
+  `baseURL` explicitly (`process.env.APP_BASE_URL || 'http://localhost:3000'`, set by
+  `containers/setup.ts` in globalSetup).
+- `npx playwright test --list` validates the whole fixture graph (scope violations, cycles,
+  load errors) **without containers or a browser** — the only meaningful local check available
+  in this sandbox. `npx tsc --noEmit -p e2e/tsconfig.json` also works, but e2e carries ~123
+  pre-existing errors (it is not covered by `npm run typecheck`), so use it differentially.
diff --git a/.claude/agent-memory/e2e-test-engineer/known-flakes-and-regressions.md b/.claude/agent-memory/e2e-test-engineer/known-flakes-and-regressions.md
index 016f4639c..902bb9655 100644
--- a/.claude/agent-memory/e2e-test-engineer/known-flakes-and-regressions.md
+++ b/.claude/agent-memory/e2e-test-engineer/known-flakes-and-regressions.md
@@ -49,6 +49,7 @@ metadata:
      Note: `invoices.spec.ts` itself has **zero diff vs `origin/beta`** — these are pre-existing latent bugs from an earlier story (#1876), not introduced by #1877; they just happened to surface in PR #1883's E2E run because full E2E runs on every PR regardless of which files changed.
 
 - **Issue #1829 (2026-07-08)** shard-3 diary flakes on main-targeted PRs: `diary-drafts.spec.ts:854` (Scenario 14) was genuinely failing on both attempt + retry — `test.slow()`'s 45s total test budget was being exhausted by 4 oversized nested step timeouts (30-45s each) before reaching cleanup, surfacing as `apiRequestContext.delete: Test timeout exceeded`. Fixed via `test.setTimeout(60_000)` + proportionate 15s step ceilings + `testInfo.setTimeout(testInfo.timeout + 15_000)` guaranteed-cleanup-time pattern in `finally`. `diary-r2-uat.spec.ts:599` (Scenario 10) had a genuine secondary race (always-empty mock defeats `waitForLoaded()`, letting a stale prior response satisfy the next `waitForResponse`) — fixed by awaiting each transition's own response explicitly and reading straight off the resolved `Response` object. `document-linking.spec.ts:369` was NOT an independent bug — confirmed via job logs it was cancelled mid-flight (1.6s runtime) as `maxFailures:1` collateral once drafts:854 exhausted its retries; no test-logic fix needed, just added the guaranteed-cleanup-time pattern defensively. Full root cause + `maxFailures`/retry-accounting proof (Playwright source citation) in [bug-1829-shard3-flakes.md](bug-1829-shard3-flakes.md). AC #3 (fail-fast retry tolerance) was already satisfied by Playwright's built-in behavior — no config change needed, only documented.
+- **Issue #1957 / PR #1961 (2026-08-02)** `no-area-filter.spec.ts:194` [mobile] "empty state when no unassigned household items exist" failed on both attempt and retry, with two `household-item-edit` timeout flakes alongside — on a PR whose entire diff was `e2e/` preference isolation and which touched neither file. **Not a flake and not a timing issue:** the PR's `test.use()` opt-ins changed worker hashes, which moved shard boundaries suite-wide, relocating `no-area-filter.spec.ts` from shard 15 into shard 14 alongside `area-filter.spec.ts` — whose Scenarios 3/5/6 hold area-less household items — and the test asserted a suite-global "no area-less item exists" precondition. Fixed by scoping the assertion to its own data (`&q=<testPrefix token>`) plus a positive control; see the pattern in [general-e2e-patterns.md](general-e2e-patterns.md) and the mechanism in [isolated-user-fixture.md](isolated-user-fixture.md). **Triage rule learned: when a shard reddens on a PR that does not touch the failing file, diff shard membership (`--list --shard=N/16`) between base and head before assuming a flake.** A green rerun proves nothing here — the collision depends on shard-internal execution order.
 - Diary Scenario 14 (`diary-drafts.spec.ts` draft-card-click-navigates) — was a persistent flake even after an earlier partial fix (PR #1671); root-caused and properly fixed via `fix/diary-scenario14-e2e-flake` (register `waitForResponse` before the click, not after). Recurred 2026-07-07 with a DIFFERENT failure signature (teardown-timeout, not nav-timing) — see #1829 entry above for the follow-up fix.
 - `invoice-budget-line-create-and-link.spec.ts` Scenarios 1–4 — REAL production regression from PR #1566 (`eagerLinkInvoice:false`), filed as bug #1611. Tests were NOT weakened; they assert correct behavior.
 - Shard 3 promotion blocker (`budget-source-filter.spec.ts`, 2026-06-12): "Rapid debounce coalesces requests" flaked on strict request-count assertions vs. CI click serialization beyond the 50ms debounce (fixed PR #1665); "Perspective toggle changes Cost value" flaked on reading `textContent()` immediately after a radio click without waiting for React re-render on WebKit (fixed PR #1666). **General fix pattern**: after any click triggering a React state update, `await expect(locator).not.toHaveText(previousValue)` before `textContent()` — never read immediately post-click on WebKit. Remove `page.on('request', ...)` listeners you add (persist for the page's lifetime otherwise); prefer state assertions (`aria-pressed`, URL params) over raw request counting.
diff --git a/e2e/fixtures/isolatedUser.ts b/e2e/fixtures/isolatedUser.ts
new file mode 100644
index 000000000..1e4ea3f95
--- /dev/null
+++ b/e2e/fixtures/isolatedUser.ts
@@ -0,0 +1,416 @@
+/**
+ * Dedicated-user isolation for preference-mutating E2E specs (Issue #1957).
+ *
+ * ─────────────────────────────────────────────────────────────────────────────
+ * WHY THIS EXISTS
+ * ─────────────────────────────────────────────────────────────────────────────
+ * `playwright.config.ts` runs with `fullyParallel: true` and every project
+ * authenticates as the SAME shared admin user (`test-results/.auth/admin.json`).
+ * Per-user preference rows (`user_preferences`, keyed by user id) are therefore a
+ * single mutable row shared by every concurrently running test in the suite.
+ *
+ * This is not merely "a test reads stale data" — it actively corrupts a running
+ * test's UI. `LocaleContext.syncWithServer` (client/src/contexts/LocaleContext.tsx)
+ * treats the SERVER value as authoritative: on every page load it fetches
+ * `/api/users/me/preferences` and, when a `locale` row exists, applies the server
+ * value AND deletes the `locale` key from `localStorage`. So if test A is asserting
+ * English text and test B (different file, different worker) PATCHes `locale='de'`
+ * on the shared admin, test A's next navigation flips to German and the
+ * `localStorage` override that would otherwise have protected it has just been
+ * deleted by the same sync call. Same class of failure for
+ * `dashboard.hiddenCards` (cards vanish/reappear mid-test) and
+ * `table.<name>.columns` (columns appear/disappear mid-test).
+ *
+ * `test.describe.configure({ mode: 'serial' })` cannot fix this: it only
+ * serializes a file against ITSELF and has no knowledge of writes coming from
+ * other files running concurrently in other workers.
+ *
+ * The remedy is to give preference-mutating tests their own disposable user, so
+ * the row they write is unreachable by any other test in the suite. This module
+ * packages that pattern (previously hand-rolled in `dashboard.spec.ts`,
+ * `i18n-categories.spec.ts` and `change-password.spec.ts`) as a fixture, so the
+ * spec bodies keep using the plain `page` fixture and keep Playwright's automatic
+ * trace/video/screenshot instrumentation (a hand-made `browser.newContext()` does
+ * not get those artifacts).
+ *
+ * ─────────────────────────────────────────────────────────────────────────────
+ * AUDIT — every spec that writes /api/users/me/preferences (Issue #1957, AC1)
+ * ─────────────────────────────────────────────────────────────────────────────
+ * Reproduce the file list with: `grep -rl "users/me/preferences" e2e/tests/`
+ * (`e2e/pages/DashboardPage.ts` and `e2e/pages/InvoicesPage.ts` also match that
+ * string, but only inside `waitForResponse()` predicates — they observe the
+ * app's own requests, they never issue one, so they are not audit entries.)
+ *
+ * 1. e2e/tests/navigation/dashboard.spec.ts
+ *    Keys: `dashboard.hiddenCards`, `locale` (top-level `beforeEach` reset, ran
+ *    for all ~34 tests) + `dashboard.hiddenCards` written by the app itself when
+ *    Scenario 6/7 click a card's dismiss/re-enable button.
+ *    Admin-gated? No. The dashboard, `/project/work-items`, `/budget/invoices`,
+ *    `/diary/new` and the Add/Customize dropdowns have no role checks. The complete
+ *    role-gating inventory in the app: server — `requireRole('admin')` on
+ *    `/api/users` mutations and `/api/backups/*`, plus work-item note
+ *    update/delete, which pass `request.user.role === 'admin'` into `noteService`
+ *    as an ownership override (`server/src/routes/notes.ts`); client — the Settings
+ *    sub-nav "User Management" and "Backups" tabs, and work-item note edit/delete
+ *    on other users' notes. Nothing else in `client/src` or `server/src` branches
+ *    on role.
+ *    Resolution (AC2): `isolatedUserPerWorker` at file scope.
+ *
+ * 2. e2e/tests/i18n/i18n.spec.ts
+ *    Keys: `locale` (PATCH in `setLanguage()`/`resetToEnglish()`, plus
+ *    `DELETE /api/users/me/preferences/locale` in "DELETE preference resets to
+ *    system locale").
+ *    Admin-gated? No — profile, dashboard, budget, schedule, diary, work items
+ *    and the Settings→Vendors tab are all member-visible.
+ *    Resolution (AC2, AC4): `isolatedUserPerWorker` at file scope. The file-level
+ *    `serial` guard added in PR #1956 is kept only as defence-in-depth against CPU
+ *    contention between two slow German-locale tests; it is no longer the
+ *    isolation mechanism.
+ *
+ * 3. e2e/tests/i18n/i18n-categories.spec.ts
+ *    Key: `locale`. ALREADY ISOLATED before this issue — creates a dedicated user
+ *    per test and logs it into its own browser context. No change needed; listed
+ *    here because AC1 requires the audit to cover every file in the grep output.
+ *
+ * 4. e2e/tests/diary/diary-uat-fixes.spec.ts
+ *    Key: `dashboard.hiddenCards` (Scenario 3 "Dashboard shows a Recent Diary
+ *    card" and Scenario 7 "Recent Diary View All link", which reset the key so
+ *    the card is guaranteed visible). Same key as dashboard.spec.ts's reset —
+ *    the second confirmed cross-file collision pair.
+ *    Admin-gated? No — both tests only read the dashboard with mocked
+ *    `/api/diary-entries`.
+ *    Resolution (AC2, AC5): `isolatedUserPerTest` at describe scope for those two
+ *    describes; the other six tests in the file touch no preferences and stay on
+ *    the shared admin.
+ *
+ * 5. e2e/tests/invoices/invoices.spec.ts
+ *    Key: `table.invoices.columns` (two DELETEs, both inside the single test
+ *    "Toggling Effective Amount ..."), plus the app's own debounced PATCH of that
+ *    key when `enableColumn()` toggles a column.
+ *    Admin-gated? No — the invoices list/detail pages and vendor/invoice/work-item/
+ *    budget-source creation have no role checks.
+ *    Resolution (AC2): `isolatedUserPerTest` at describe scope for the
+ *    "Effective Amount" describe only.
+ *
+ * No audited spec needs the admin role, so no spec required the AC3 treatment.
+ * `role: 'admin'` is supported below for any future case that does.
+ *
+ * Related shared-state hazards deliberately OUT of scope here (they do not write
+ * `/api/users/me/preferences` and so are outside AC1's grep, but the next person
+ * should know): the app persists `table.<name>.columns` for every DataTable, so
+ * any spec that toggles columns as the shared admin writes a preference row for
+ * that table — only a spec asserting the same table's columns can be a victim,
+ * and `invoices.spec.ts` is currently the only such spec. Dark-mode specs set
+ * `data-theme` on the document instead of persisting a preference, so they are
+ * not affected.
+ *
+ * ─────────────────────────────────────────────────────────────────────────────
+ * USAGE
+ * ─────────────────────────────────────────────────────────────────────────────
+ * File scope — one dedicated user per worker, shared by every test in the file:
+ *
+ *   import { test, expect } from '../../fixtures/isolatedUser.js';
+ *   test.use({ isolatedUserPerWorker: { emailPrefix: 'dash' } });
+ *
+ * Describe scope — a fresh dedicated user for each test in that describe:
+ *
+ *   test.describe('...', () => {
+ *     test.use({ isolatedUserPerTest: { emailPrefix: 'diary-dash' } });
+ *   });
+ *
+ * Either way the built-in `page`/`context` fixtures are already authenticated as
+ * the dedicated user, so `page.request.patch('/api/users/me/preferences', ...)`
+ * inside a test writes that user's row and nothing else's. Tests that need the
+ * user's identity can read the `isolatedUserSession` fixture.
+ *
+ * WHICH SCOPE TO PICK
+ * - `isolatedUserPerWorker` is the default choice. A Playwright worker runs one
+ *   test at a time, and no other worker shares the user, so the row still cannot
+ *   be written concurrently by anything else in the suite — while costing one user
+ *   per worker instead of one per test. State CAN carry over between the
+ *   sequential tests of one worker, so a file using it must still reset any key it
+ *   dirties (see dashboard.spec.ts's `beforeEach` / i18n.spec.ts's `afterEach`).
+ * - `isolatedUserPerTest` gives a guaranteed-pristine row per test, at the cost of
+ *   one user creation + login per test (~0.5-1s, which counts against the test
+ *   timeout). Use it for a handful of tests inside an otherwise shared-admin file,
+ *   because a worker-scoped option cannot be set from inside a `describe`
+ *   (Playwright rejects it: "Cannot use({...}) in a describe group, because it
+ *   forces a new worker").
+ *
+ * !! EXPECT AN UNRELATED SHARD TO GO RED WHEN YOU OPT A FILE IN !!
+ * Adding `test.use()` for either option changes the affected tests' `_workerHash`.
+ * `createTestGroups` (node_modules/playwright/lib/runner/index.js) buckets tests by
+ * worker hash FIRST and emits groups in bucket insertion order; `filterForShard`
+ * then slices that list by cumulative test count. So opting one file in
+ * redistributes shard MEMBERSHIP across the whole suite — not just this file's —
+ * even when the total test count is unchanged. Any latent cross-file shared-state
+ * assumption in a newly co-located pair surfaces as a failure that looks unrelated
+ * to your change. #1957 hit exactly this: `no-area-filter.spec.ts` moved from shard
+ * 15 to shard 14 next to `area-filter.spec.ts` and its suite-global "no area-less
+ * household item exists" precondition broke. Diagnose before blaming the change
+ * itself, by diffing shard membership between base and head:
+ *   git archive <base-sha> e2e | tar -x -C /tmp/base && ln -s <repo>/node_modules /tmp/base/node_modules
+ *   (cd /tmp/base && npx playwright test --list --shard=N/16)   # vs the same in the worktree
+ * Do NOT respond by pinning or reordering shards (#1957's Notes rule that out) —
+ * fix the test that assumes suite-global state.
+ *
+ * Users are created via `POST /api/users` (admin-only) and removed in fixture
+ * teardown via `DELETE /api/users/:id`. Note that DELETE is a SOFT delete
+ * (`deactivateUser` + `destroyUserSessions`): the preference row survives, but the
+ * account can never be logged into again and its e-mail is never reused, so
+ * nothing can read or write that row afterwards. Accumulation is bounded per
+ * SHARD, not per run — each shard is its own CI job with its own container and DB
+ * (`containers/setup.ts` runs in globalSetup) — so `/settings/users`' 100-row page
+ * is never in reach. The reason to prefer per-worker is the login rate limit
+ * (`POST /api/auth/login`: 20 requests / 15 min, keyed on `request.ip`, one bucket
+ * per shard), not user-table size.
+ */
+
+import type { APIRequestContext, PlaywrightWorkerArgs } from '@playwright/test';
+import { test as authTest } from './auth.js';
+import { API } from './testData.js';
+
+/**
+ * Shared-admin storage state written by `auth.setup.ts`; mirrors `use.storageState`
+ * in playwright.config.ts. Used here to authenticate the admin API context that
+ * provisions and deactivates dedicated users. (Non-opted-in tests get their state
+ * from `testInfo.project.use.storageState`, not from this constant.)
+ */
+export const ADMIN_STORAGE_STATE = 'test-results/.auth/admin.json';
+
+/** Password given to every dedicated user (>= 8 chars per createUserSchema). */
+const ISOLATED_USER_PASSWORD = 'e2e-isolated-pw-123!';
+
+/** Opt-in configuration for a dedicated user. */
+export interface IsolatedUserSpec {
+  /**
+   * Short slug used in the generated e-mail address, for readability in failure
+   * output. Keep the string "admin" out of it — `search-users.spec.ts` asserts
+   * that every row matching a search for "Admin" has "admin" in its name column.
+   */
+  emailPrefix: string;
+  /** Defaults to 'E2E Isolated User'. Must not contain "admin" (see above). */
+  displayName?: string;
+  /** Defaults to 'member'. Only set 'admin' for an actual admin-gated dependency. */
+  role?: 'admin' | 'member';
+  /**
+   * `locale` preference to seed on the dedicated user. Defaults to 'en' so specs
+   * asserting English strings do not depend on the CI browser's default locale.
+   */
+  locale?: 'en' | 'de' | 'system';
+}
+
+/** Identity of the dedicated user backing the current test's browser context. */
+export interface IsolatedUserSession {
+  id: string;
+  email: string;
+  password: string;
+  storageState: StorageStateValue;
+}
+
+type StorageStateValue = Awaited<ReturnType<APIRequestContext['storageState']>>;
+
+/**
+ * Creates a dedicated user, logs it in, seeds its `locale` preference and returns
+ * both its identity and a storage state carrying its session cookie.
+ *
+ * `baseURL` is passed in explicitly: Playwright only merges the project's context
+ * options (including `baseURL`) into contexts created *inside a test*, and a
+ * worker-scoped fixture runs before that instrumentation is installed.
+ */
+async function provisionIsolatedUser(
+  playwright: PlaywrightWorkerArgs['playwright'],
+  baseURL: string,
+  spec: IsolatedUserSpec,
+  uniqueSuffix: string,
+): Promise<{ session: IsolatedUserSession; dispose: () => Promise<void> }> {
+  const adminApi = await playwright.request.newContext({
+    baseURL,
+    storageState: ADMIN_STORAGE_STATE,
+  });
+
+  const email = `${spec.emailPrefix}-${uniqueSuffix}@e2e-test.local`;
+  let userId: string | null = null;
+
+  try {
+    const createResponse = await adminApi.post(API.users, {
+      data: {
+        email,
+        displayName: spec.displayName ?? 'E2E Isolated User',
+        password: ISOLATED_USER_PASSWORD,
+        role: spec.role ?? 'member',
+      },
+    });
+    if (!createResponse.ok()) {
+      throw new Error(
+        `isolatedUser: POST ${API.users} for "${email}" failed with ${createResponse.status()}: ${await createResponse.text()}`,
+      );
+    }
+    const created = (await createResponse.json()) as { user: { id: string } };
+    userId = created.user.id;
+
+    const userApi = await playwright.request.newContext({ baseURL });
+    let storageState: StorageStateValue;
+    try {
+      const loginResponse = await userApi.post(API.login, {
+        data: { email, password: ISOLATED_USER_PASSWORD },
+      });
+      if (!loginResponse.ok()) {
+        throw new Error(
+          `isolatedUser: login as "${email}" failed with ${loginResponse.status()}: ${await loginResponse.text()}`,
+        );
+      }
+      const localeResponse = await userApi.patch('/api/users/me/preferences', {
+        data: { key: 'locale', value: spec.locale ?? 'en' },
+      });
+      if (!localeResponse.ok()) {
+        throw new Error(
+          `isolatedUser: seeding locale for "${email}" failed with ${localeResponse.status()}`,
+        );
+      }
+      storageState = await userApi.storageState();
+    } finally {
+      // The session lives on the server; disposing the client is safe.
+      await userApi.dispose();
+    }
+
+    return {
+      session: { id: userId, email, password: ISOLATED_USER_PASSWORD, storageState },
+      dispose: async () => {
+        // Soft-delete (deactivate) the user and drop its sessions, then release
+        // the admin client. Failures are non-fatal: the account is never reused.
+        await adminApi.delete(`${API.users}/${userId}`);
+        await adminApi.dispose();
+      },
+    };
+  } catch (error) {
+    if (userId) await adminApi.delete(`${API.users}/${userId}`);
+    await adminApi.dispose();
+    throw error;
+  }
+}
+
+/**
+ * Base URL of the app under test. Mirrors playwright.config.ts's
+ * `use.baseURL`; `containers/setup.ts` sets APP_BASE_URL to the proxy URL in
+ * globalSetup, which worker processes inherit. Needed because a worker-scoped
+ * fixture cannot depend on the test-scoped `baseURL` fixture.
+ */
+function resolveBaseURL(): string {
+  return process.env.APP_BASE_URL || 'http://localhost:3000';
+}
+
+export const test = authTest.extend<
+  {
+    isolatedUserPerTest: IsolatedUserSpec | null;
+    isolatedUserSession: IsolatedUserSession | null;
+    isolatedUserGuard: void;
+  },
+  {
+    isolatedUserPerWorker: IsolatedUserSpec | null;
+    workerIsolatedUserSession: IsolatedUserSession | null;
+  }
+>({
+  // ── Options ───────────────────────────────────────────────────────────────
+  isolatedUserPerWorker: [null, { scope: 'worker', option: true }],
+  isolatedUserPerTest: [null, { option: true }],
+
+  // ── One dedicated user per worker (file-scope opt-in) ──────────────────────
+  workerIsolatedUserSession: [
+    async ({ playwright, isolatedUserPerWorker }, use, workerInfo) => {
+      if (!isolatedUserPerWorker) {
+        await use(null);
+        return;
+      }
+      const { session, dispose } = await provisionIsolatedUser(
+        playwright,
+        resolveBaseURL(),
+        isolatedUserPerWorker,
+        `${workerInfo.project.name}-w${workerInfo.workerIndex}-${Date.now()}`,
+      );
+      try {
+        await use(session);
+      } finally {
+        await dispose();
+      }
+    },
+    { scope: 'worker' },
+  ],
+
+  // ── The session backing this test's browser context ───────────────────────
+  isolatedUserSession: async (
+    { playwright, baseURL, isolatedUserPerTest, workerIsolatedUserSession },
+    use,
+    testInfo,
+  ) => {
+    if (isolatedUserPerTest && workerIsolatedUserSession) {
+      throw new Error(
+        'isolatedUser: set either isolatedUserPerWorker (file scope) or isolatedUserPerTest (file/describe scope), not both.',
+      );
+    }
+    if (!isolatedUserPerTest) {
+      await use(workerIsolatedUserSession);
+      return;
+    }
+    const { session, dispose } = await provisionIsolatedUser(
+      playwright,
+      baseURL ?? resolveBaseURL(),
+      isolatedUserPerTest,
+      `${testInfo.project.name}-w${testInfo.workerIndex}-${Date.now()}`,
+    );
+    try {
+      await use(session);
+    } finally {
+      await dispose();
+    }
+  },
+
+  // ── Point the built-in context/page fixtures at that session ──────────────
+  //
+  // The `testInfo.project.use.storageState` fallback is load-bearing, not
+  // defensive: playwright.config.ts sets `storageState` per project, and once the
+  // option is overridden here the config value is no longer consulted — without it,
+  // non-opted-in tests in an importing file would lose admin auth entirely.
+  // Deliberately resolves to `undefined` rather than forcing ADMIN_STORAGE_STATE if
+  // a project ever sets no storageState, so a future unauthenticated project keeps
+  // its intended state.
+  storageState: async ({ isolatedUserSession }, use, testInfo) => {
+    await use(isolatedUserSession?.storageState ?? testInfo.project.use.storageState);
+  },
+
+  // ── Fail loudly if the override above ever stops taking effect ────────────
+  //
+  // The whole point of this module is that `page` belongs to the dedicated user.
+  // If that silently regressed (e.g. a Playwright change to how the `storageState`
+  // option feeds `_combinedContextOptions`), the specs would quietly go back to
+  // mutating the shared admin and the collisions this fixture prevents would
+  // return unnoticed — every test would still pass. One cheap request per test
+  // turns that into an immediate, explicit failure.
+  //
+  // Two limits, both deliberate:
+  //   (a) It validates the `page` fixture only. A context a test builds itself via
+  //       `browser.newContext()` is outside its reach — such a context inherits
+  //       this override unless it passes an explicit `storageState` (as
+  //       invoices.spec.ts's Dark mode describe does, which is why that
+  //       non-converted describe is unaffected).
+  //   (b) It is `auto` and depends on `page`, so every test in an importing file
+  //       gets a browser context — including one a `beforeEach` immediately
+  //       `test.skip()`s, since fixtures resolve before hooks run.
+  isolatedUserGuard: [
+    async ({ page, isolatedUserSession }, use) => {
+      if (isolatedUserSession) {
+        const response = await page.request.get(API.authMe);
+        const body = (await response.json()) as { user: { email: string } | null };
+        if (body.user?.email !== isolatedUserSession.email) {
+          throw new Error(
+            `isolatedUser: expected the test's browser context to be authenticated as "${isolatedUserSession.email}", but /api/auth/me reports "${body.user?.email ?? 'nobody'}". The storageState override is not taking effect — preference writes would hit the shared admin user.`,
+          );
+        }
+      }
+      await use();
+    },
+    { auto: true },
+  ],
+});
+
+export { expect } from '@playwright/test';
diff --git a/e2e/tests/diary/diary-uat-fixes.spec.ts b/e2e/tests/diary/diary-uat-fixes.spec.ts
index 591380a96..38f5918f6 100644
--- a/e2e/tests/diary/diary-uat-fixes.spec.ts
+++ b/e2e/tests/diary/diary-uat-fixes.spec.ts
@@ -21,7 +21,7 @@
  * 8.  Diary detail page has no print button
  */
 
-import { test, expect } from '../../fixtures/auth.js';
+import { test, expect } from '../../fixtures/isolatedUser.js';
 import { DiaryPage, DIARY_ROUTE } from '../../pages/DiaryPage.js';
 import { DiaryEntryDetailPage } from '../../pages/DiaryEntryDetailPage.js';
 import { DiaryEntryCreatePage } from '../../pages/DiaryEntryCreatePage.js';
@@ -82,6 +82,16 @@ test.describe('Back button navigates to /diary (Scenario 2)', { tag: '@responsiv
 // Scenario 3: Dashboard "Recent Diary" card is visible
 // ─────────────────────────────────────────────────────────────────────────────
 test.describe('Dashboard Recent Diary card (Scenario 3)', { tag: '@responsive' }, () => {
+  // This test resets `dashboard.hiddenCards` so the Recent Diary card is guaranteed
+  // to render. On the shared admin that reset collided with dashboard.spec.ts, which
+  // resets the same key for its own tests and asserts card visibility against it
+  // (Issue #1957). Run it as a dedicated user instead, so the row is unreachable by
+  // any other test. Per-test rather than per-worker because a worker-scoped option
+  // cannot be set inside a describe, and only 2 of this file's 8 tests need it.
+  // Nothing here is admin-gated: the test only reads the dashboard with
+  // `/api/diary-entries` mocked.
+  test.use({ isolatedUserPerTest: { emailPrefix: 'diary-dash', displayName: 'E2E Diary User' } });
+
   test('Dashboard page shows a "Recent Diary" card', { tag: '@smoke' }, async ({ page }) => {
     const dashboardPage = new DashboardPage(page);
 
@@ -102,7 +112,10 @@ test.describe('Dashboard Recent Diary card (Scenario 3)', { tag: '@responsive' }
     });
 
     try {
-      // Reset hidden cards to ensure "Recent Diary" is visible
+      // Reset hidden cards on this test's dedicated user to ensure "Recent Diary" is
+      // visible. A freshly created user has no preference rows at all, so this is
+      // belt-and-braces rather than load-bearing — but it keeps the precondition
+      // explicit and would still hold if a dismiss step were added here later.
       await page.request.patch('/api/users/me/preferences', {
         data: { key: 'dashboard.hiddenCards', value: '[]' },
       });
@@ -322,6 +335,9 @@ test.describe('Create navigates to detail page (Scenario 6)', { tag: '@responsiv
 // Scenario 7: Dashboard diary card "View All" navigates to /diary
 // ─────────────────────────────────────────────────────────────────────────────
 test.describe('Recent Diary "View All" link (Scenario 7)', { tag: '@responsive' }, () => {
+  // Same `dashboard.hiddenCards` collision as Scenario 3 — see the comment there.
+  test.use({ isolatedUserPerTest: { emailPrefix: 'diary-dash', displayName: 'E2E Diary User' } });
+
   test('Clicking "View All" in the Recent Diary card navigates to /diary', async ({ page }) => {
     const dashboardPage = new DashboardPage(page);
 
@@ -360,7 +376,7 @@ test.describe('Recent Diary "View All" link (Scenario 7)', { tag: '@responsive'
     });
 
     try {
-      // Reset hidden cards
+      // Reset hidden cards on this test's dedicated user (see Scenario 3's note)
       await page.request.patch('/api/users/me/preferences', {
         data: { key: 'dashboard.hiddenCards', value: '[]' },
       });
diff --git a/e2e/tests/household-items/no-area-filter.spec.ts b/e2e/tests/household-items/no-area-filter.spec.ts
index c120da295..1c81ad67b 100644
--- a/e2e/tests/household-items/no-area-filter.spec.ts
+++ b/e2e/tests/household-items/no-area-filter.spec.ts
@@ -15,7 +15,9 @@
  * 1.  Sentinel renders at top of popover (desktop/tablet; mobile skip)
  * 2.  ?areaId=__none__ shows only unassigned items
  * 3.  ?areaId=__none__,<areaId> shows union of unassigned + named area items
- * 4.  Empty state when no unassigned items exist and ?areaId=__none__ applied
+ * 4.  Empty state when none of this test's own items are unassigned and
+ *     ?areaId=__none__ applied — see the note on Scenario 4 below for why the
+ *     assertion is scoped with `&q=` instead of assuming a suite-global state.
  */
 
 import { test, expect } from '../../fixtures/auth.js';
@@ -183,15 +185,41 @@ test.describe(
 );
 
 // ─────────────────────────────────────────────────────────────────────────────
-// Scenario 4: Empty state when no unassigned items and ?areaId=__none__ applied
+// Scenario 4: Empty state when none of this test's items are unassigned and
+// ?areaId=__none__ applied
+//
+// This assertion must NOT be phrased as "no unassigned household item exists
+// anywhere", which is what it used to do: `?areaId=__none__` alone lists every
+// area-less item in the shared DB, and 48 of the suite's 62
+// `createHouseholdItemViaApi()` calls — spread over 12 spec files — pass no areaId
+// (area-filter.spec.ts:185, :318 and :377 each hold one for the length of a long
+// scenario). Under `fullyParallel: true` a single foreign area-less item makes the
+// empty state unreachable, so the test only ever passed while no such spec happened
+// to share its shard — it broke the moment #1957's worker-hash change moved this
+// file from shard 15 into shard 14, whose other members are area-filter.spec.ts,
+// household-items-list.spec.ts, household-item-create/-detail/-edit.spec.ts.
+//
+// The sibling `e2e/tests/work-items/no-area-filter.spec.ts:223` already solves this
+// the same way and has been green in shards 7/12/16 while co-resident with five
+// area-less work-item creators — this file was simply never given the same
+// treatment.
+//
+// Fix: AND the sentinel filter with a `q=` search for this scenario's own name
+// token (`useTableState` hydrates `q` from the URL and householdItemService ANDs
+// it with the areaId condition), so only items this test created can satisfy the
+// list. `q` also counts towards DataTable's `hasActiveFilters`, so the filtered
+// empty-state message and Clear Filters button are still the correct expectations.
+// A positive control runs first: the same `q` without the sentinel must list the
+// area-assigned item, which rules out the empty state passing vacuously because
+// the search matched nothing at all.
 // ─────────────────────────────────────────────────────────────────────────────
 test.describe(
-  '?areaId=__none__ shows filtered empty state when no unassigned household items exist (Scenario 4)',
+  '?areaId=__none__ shows filtered empty state when no matching household item is unassigned (Scenario 4)',
   { tag: '@responsive' },
   () => {
     test.describe.configure({ timeout: 90_000 });
 
-    test('Empty state with Clear Filters button when all items are area-assigned', async ({
+    test('Empty state with Clear Filters button when the only matching item is area-assigned', async ({
       page,
       testPrefix,
     }) => {
@@ -199,16 +227,43 @@ test.describe(
       const areaIds: string[] = [];
       const itemIds: string[] = [];
 
-      const areaName = `${testPrefix} HI NoArea Sc4 Area`;
-      const itemAssignedName = `${testPrefix} HI NoArea Sc4 Assigned`;
+      // Search token that scopes every assertion below to this test's own data.
+      // Unique per worker+project via testPrefix, and no other test uses this
+      // scenario suffix.
+      const scopeToken = `${testPrefix} HI NoArea Sc4`;
+      const areaName = `${scopeToken} Area`;
+      const itemAssignedName = `${scopeToken} Assigned`;
 
       try {
         const areaId = await createAreaViaApi(page, { name: areaName });
         areaIds.push(areaId);
         itemIds.push(await createHouseholdItemViaApi(page, { name: itemAssignedName, areaId }));
 
-        await page.goto(`${HOUSEHOLD_ITEMS_ROUTE}?areaId=__none__`);
+        // Positive control: the search token alone must find the area-assigned item.
+        // Without this, an empty result below would be indistinguishable from "the
+        // search matched nothing", and the empty-state assertion would pass even if
+        // the sentinel filter did nothing.
+        await page.goto(`${HOUSEHOLD_ITEMS_ROUTE}?q=${encodeURIComponent(scopeToken)}`);
         await listPage.heading.waitFor({ state: 'visible' });
+        await listPage.waitForLoaded();
+        await expect(async () => {
+          const names = await listPage.getItemNames();
+          expect(names).toContain(itemAssignedName);
+        }).toPass({ timeout: 30_000 });
+
+        // Now add the sentinel: the one matching item has an area, so the filtered
+        // result must be empty regardless of what other specs are doing concurrently.
+        await page.goto(
+          `${HOUSEHOLD_ITEMS_ROUTE}?areaId=__none__&q=${encodeURIComponent(scopeToken)}`,
+        );
+        await listPage.heading.waitFor({ state: 'visible' });
+
+        // Both filters must survive the SPA's URL round-trip (handleStateChange
+        // rewrites the query string from table state), otherwise the empty state
+        // below could be caused by a different filter set than the one under test.
+        const url = new URL(page.url());
+        expect(url.searchParams.get('areaId')).toBe('__none__');
+        expect(url.searchParams.get('q')).toBe(scopeToken);
 
         // Wait for the filter empty state
         await expect(async () => {
diff --git a/e2e/tests/i18n/i18n-categories.spec.ts b/e2e/tests/i18n/i18n-categories.spec.ts
index 92182eed3..205bd24b0 100644
--- a/e2e/tests/i18n/i18n-categories.spec.ts
+++ b/e2e/tests/i18n/i18n-categories.spec.ts
@@ -9,7 +9,11 @@
  * Test strategy:
  * - Each test creates a dedicated local user so that PATCH /api/users/me/preferences
  *   never mutates the shared TEST_ADMIN user, eliminating locale-state leakage across
- *   parallel workers.
+ *   parallel workers. (This file is listed in the preference-write audit in
+ *   e2e/fixtures/isolatedUser.ts, Issue #1957, as already-isolated. The hand-rolled
+ *   helper below predates that fixture and is equivalent to
+ *   `test.use({ isolatedUserPerTest: ... })`; it is left as-is because it also needs
+ *   to write localStorage on the scoped page before the first navigation.)
  * - A fresh browser context (no storageState) is created per test, logged in as the
  *   dedicated user, and closed in a finally block.
  * - Scoped to desktop only — language state changes involve localStorage + API
diff --git a/e2e/tests/i18n/i18n.spec.ts b/e2e/tests/i18n/i18n.spec.ts
index 24e587d36..dfdec5775 100644
--- a/e2e/tests/i18n/i18n.spec.ts
+++ b/e2e/tests/i18n/i18n.spec.ts
@@ -16,28 +16,38 @@
  * Scoped to desktop only: language switching involves a form that can be unreliable
  * on WebKit tablet — and language correctness is not viewport-specific.
  *
- * Serial mode (file scope): every test here mutates the SAME row —
- * user_preferences(locale) of the shared TEST_ADMIN user. Under `fullyParallel: true`
- * two of these tests run concurrently in different workers, so one test's
- * `setLanguage()`/`resetToEnglish()` PATCH lands inside another's assertions. That is
- * not merely "stale data": LocaleContext.syncWithServer treats the server value as
- * authoritative, so it applies the other test's locale AND deletes the victim's
- * 'locale' localStorage key, permanently flipping the victim's UI language mid-test.
- * Observed in CI run 30790367863 shard 4, where "Key page headings render in German"
- * asserted 'Projekt' successfully at 06:35:35.49 and then found an English sidebar
- * ('Main navigation', 'Schedule') at 06:35:36.11 — the window in which the
- * concurrently running "Language can be switched back to English from German"
- * (06:35:34.41–35.88, worker 1) PATCHed locale='en'.
- * There is only one admin user, so `testPrefix` cannot isolate this; serial mode is
- * the established remedy for shared-admin-mutating specs (cf. change-password.spec.ts,
- * edit-user.spec.ts). It must be file-scoped, not describe-scoped: the interference
- * observed above crossed describe boundaries.
+ * Preference isolation (Issue #1957): every test here mutates the `locale`
+ * preference row of whichever user it is authenticated as. That used to be the one
+ * shared TEST_ADMIN user, which under `fullyParallel: true` made this file both a
+ * victim and a cause of cross-file corruption — LocaleContext.syncWithServer treats
+ * the server value as authoritative, so a concurrent PATCH from another worker
+ * applies its locale AND deletes the victim's 'locale' localStorage key, flipping
+ * the victim's UI language mid-test. Observed in CI run 30790367863 shard 4, where
+ * "Key page headings render in German" asserted 'Projekt' successfully at
+ * 06:35:35.49 and then found an English sidebar ('Main navigation', 'Schedule') at
+ * 06:35:36.11 — the window in which the concurrently running "Language can be
+ * switched back to English from German" (06:35:34.41–35.88, worker 1) PATCHed
+ * locale='en'. `dashboard.spec.ts` was the mirror-image victim: its top-level
+ * `beforeEach` PATCHed locale='en' on the same shared row.
+ *
+ * The isolation mechanism is now `isolatedUserPerWorker` (see
+ * e2e/fixtures/isolatedUser.ts): every test runs as a dedicated user, so the row
+ * `setLanguage()` writes is unreachable by any other test in the suite. Serial mode
+ * is KEPT, but only as defence in depth: it no longer provides isolation (it never
+ * could across files), it just avoids running two slow German cold-start tests
+ * concurrently on a 2-vCPU CI runner. `resetToEnglish()` in afterEach is also kept
+ * and is still load-bearing — a worker's dedicated user is shared by that worker's
+ * sequentially-executed tests, so locale state must not carry over between them.
  */
 
 import type { Page } from '@playwright/test';
-import { test, expect } from '../../fixtures/auth.js';
+import { test, expect } from '../../fixtures/isolatedUser.js';
 import { ROUTES } from '../../fixtures/testData.js';
 
+test.use({
+  isolatedUserPerWorker: { emailPrefix: 'i18n-switch', displayName: 'E2E i18n User' },
+});
+
 test.describe.configure({ mode: 'serial' });
 
 // ─────────────────────────────────────────────────────────────────────────────
diff --git a/e2e/tests/invoices/invoices.spec.ts b/e2e/tests/invoices/invoices.spec.ts
index c133d3855..a9ca9c457 100644
--- a/e2e/tests/invoices/invoices.spec.ts
+++ b/e2e/tests/invoices/invoices.spec.ts
@@ -19,7 +19,7 @@
  * - Dark mode
  */
 
-import { test, expect } from '../../fixtures/auth.js';
+import { test, expect } from '../../fixtures/isolatedUser.js';
 import type { Page } from '@playwright/test';
 import { InvoicesPage } from '../../pages/InvoicesPage.js';
 import { InvoiceDetailPage } from '../../pages/InvoiceDetailPage.js';
@@ -838,6 +838,19 @@ test.describe('Dark mode', () => {
 // ─────────────────────────────────────────────────────────────────────────────
 
 test.describe('"Effective Amount" column (Issue #1876)', { tag: '@responsive' }, () => {
+  // This is the only test in the suite that asserts against `table.invoices.columns`,
+  // a per-user singleton preference that `useColumnPreferences` writes (debounced) for
+  // every visitor of the invoices table. On the shared admin, any concurrently running
+  // test that renders that table could move the column set under this test's feet, and
+  // this test's own DELETEs could equally disturb others (Issue #1957). Run it as a
+  // dedicated user so the row belongs to nobody else. Per-test rather than per-worker
+  // because a worker-scoped option cannot be set inside a describe. Nothing here is
+  // admin-gated: the invoices list/detail pages and vendor/invoice/work-item/
+  // budget-source creation have no role checks.
+  test.use({
+    isolatedUserPerTest: { emailPrefix: 'inv-columns', displayName: 'E2E Invoices User' },
+  });
+
   test('Toggling "Effective Amount" shows a deposit/refund-aware value distinct from "Remaining Amount"', async ({
     page,
     testPrefix,
@@ -908,9 +921,10 @@ test.describe('"Effective Amount" column (Issue #1876)', { tag: '@responsive' },
       // Effective Amount = 1000 − 150 (paid refund) = 850
 
       // Reset the "table.invoices.columns" preference before asserting the
-      // hidden-by-default baseline — a prior run's debounced save (or this test's
-      // own retry) can otherwise leave "Effective Amount"/"Remaining Amount"
-      // already-visible for this account before we even start.
+      // hidden-by-default baseline. Since this test now runs as a freshly created
+      // dedicated user (see test.use above), the row cannot pre-exist — not even on a
+      // retry, which provisions a new user. Kept as an explicit precondition so the
+      // baseline assertion below cannot silently depend on account history.
       await page.request.delete('/api/users/me/preferences/table.invoices.columns');
 
       await invoicesPage.goto();
@@ -964,10 +978,10 @@ test.describe('"Effective Amount" column (Issue #1876)', { tag: '@responsive' },
     } finally {
       // Column visibility/order is a per-user server-side SINGLETON preference
       // (`table.invoices.columns`, see useColumnPreferences), not a per-test entity.
-      // Reset it so this test's "enable Remaining/Effective Amount" toggles never
-      // leak into a retry of this same test (the debounced save can persist after a
-      // failed assertion) or into any other invoices test running afterward against
-      // the same account. DELETE 404s if no preference was ever saved — fine either way.
+      // The dedicated user is deactivated in fixture teardown, so these toggles can no
+      // longer leak anywhere; the reset is kept so the intent survives if this describe
+      // is ever switched back to a shared account.
+      // DELETE 404s if no preference was ever saved — fine either way.
       await page.request.delete('/api/users/me/preferences/table.invoices.columns');
       if (vendorId) await deleteVendorViaApi(page, vendorId);
       if (workItemId) await deleteWorkItemViaApi(page, workItemId);
diff --git a/e2e/tests/navigation/dashboard.spec.ts b/e2e/tests/navigation/dashboard.spec.ts
index 066912537..f14f95a63 100644
--- a/e2e/tests/navigation/dashboard.spec.ts
+++ b/e2e/tests/navigation/dashboard.spec.ts
@@ -17,17 +17,45 @@
  * 11. No horizontal scroll on current viewport
  */
 
-import type { Browser, BrowserContext, Page } from '@playwright/test';
-import { test, expect } from '../../fixtures/auth.js';
+import { test, expect } from '../../fixtures/isolatedUser.js';
 import { DashboardPage, DASHBOARD_ROUTE, CARD_TITLES } from '../../pages/DashboardPage.js';
-import { createLocalUserViaApi, deleteUserViaApi } from '../../fixtures/apiHelpers.js';
 
 // ─────────────────────────────────────────────────────────────────────────────
-// Global setup: reset dashboard preferences before every test to prevent
-// state leaking from dismiss tests (Issue 1: server-side preference persistence)
+// Preference isolation (Issue #1957)
+//
+// Every test in this file depends on two per-user preference rows:
+// `dashboard.hiddenCards` (which cards render) and `locale` (English card
+// headings). Scenario 6/7 additionally WRITE `dashboard.hiddenCards` by clicking
+// dismiss/re-enable, and the reset hook below writes both keys.
+//
+// Under `fullyParallel: true` all of that used to happen on the one shared admin
+// user (test-results/.auth/admin.json), so any other spec touching those keys —
+// diary-uat-fixes.spec.ts resets `dashboard.hiddenCards`, i18n.spec.ts flips
+// `locale` — could land a write inside a test's assertion window from another
+// worker, and this file's own reset hook could wipe out Scenario 6's dismissed
+// state mid-test. `mode: 'serial'` cannot fix that: it only orders a file against
+// itself.
+//
+// This file therefore runs against a dedicated user (see e2e/fixtures/isolatedUser.ts
+// for the full mechanism and the audit of every preference-writing spec). One
+// dedicated user per worker is enough: a Playwright worker executes one test at a
+// time and no other worker shares the user, so no concurrent write to those rows
+// is possible from anywhere in the suite. Sequential carry-over inside one worker
+// is still possible (Scenario 6 leaves a card hidden for whatever test runs next
+// in that worker), which is exactly what the reset hook below handles.
+//
+// The write/read path itself is correctly ordered and durable (dismissCard() awaits
+// the PATCH response; preferencesService.upsertPreference() commits synchronously) —
+// this was a test-isolation gap, never a product bug.
 // ─────────────────────────────────────────────────────────────────────────────
 
+test.use({
+  isolatedUserPerWorker: { emailPrefix: 'dash', displayName: 'E2E Dashboard User' },
+});
+
 test.beforeEach(async ({ page }) => {
+  // Reset this worker's dedicated user back to "no cards hidden" so a preceding
+  // dismiss test in the same worker cannot leak into the next one.
   const resp = await page.request.patch('/api/users/me/preferences', {
     data: { key: 'dashboard.hiddenCards', value: '[]' },
   });
@@ -35,9 +63,10 @@ test.beforeEach(async ({ page }) => {
   // from a prior test, causing downstream dismiss tests to fail.
   expect(resp.ok(), `beforeEach: preference reset failed with ${resp.status()}`).toBeTruthy();
 
-  // Also reset the locale preference to English. If an i18n test in the same shard
-  // left the locale as 'de', the dashboard would render with German card headings
-  // (e.g., "Schnellaktionen" instead of "Quick Actions"), causing locator failures.
+  // Force English regardless of the CI browser's default locale — every assertion
+  // in this file matches English headings (e.g. "Quick Actions", not
+  // "Schnellaktionen"). The dedicated user is seeded with locale='en' on creation;
+  // this re-asserts it in case a test in this worker changed it.
   await page.request.patch('/api/users/me/preferences', {
     data: { key: 'locale', value: 'en' },
   });
@@ -480,82 +509,22 @@ test.describe('Quick Actions card (Scenario 5)', { tag: '@responsive' }, () => {
   });
 });
 
-// ─────────────────────────────────────────────────────────────────────────────
-// Isolated-user helper for Scenario 6 & 7 (card dismiss / re-enable)
-//
-// Root cause of the intermittent "Dismissed card stays hidden after page reload"
-// failure (PR #1935, shard 6): `dashboard.hiddenCards` is a single preference row
-// keyed by userId. Every test in this file authenticates as the one shared admin
-// user (test-results/.auth/admin.json), and playwright.config.ts runs with
-// `fullyParallel: true`, so unrelated tests/describe-blocks in this same file
-// (each of which runs the top-level `beforeEach` that PATCHes
-// `dashboard.hiddenCards` back to "[]" on that same shared admin user) can be
-// scheduled onto a different worker and land their reset in the narrow window
-// between this test's own dismiss-PATCH and its post-reload GET — wiping out the
-// very state under test. The previous `mode: 'serial'` guard only serialized
-// Scenario 6's own two tests against each other; it could not stop the other
-// ~30 tests in this file from touching the same shared-admin preference row
-// concurrently, since they have no knowledge of Scenario 6's serial group.
-//
-// The write/read path itself (upsertPreference -> synchronous better-sqlite3
-// write -> 200 response -> dismissCard() awaits that response -> reload -> GET)
-// is correctly ordered and durable — there is no persistence bug in production
-// code here (see `dismissCard()` in DashboardPage.ts and
-// `preferencesService.upsertPreference()`, both of which resolve/commit before
-// the next step runs). This was a test-isolation gap, not a product bug.
-//
-// Fix: give each dismiss/re-enable test its own dedicated user + browser
-// context, mirroring the established pattern in
-// e2e/tests/i18n/i18n-categories.spec.ts and e2e/tests/profile/change-password.spec.ts.
-// A dedicated user's `dashboard.hiddenCards` row can never be touched by any
-// other concurrently running test, so the persistence contract can be verified
-// without racing the rest of the suite. `ON DELETE CASCADE` on
-// user_preferences.user_id means deleting the user also removes its
-// preferences — no separate reset/cleanup PATCH is needed.
-// ─────────────────────────────────────────────────────────────────────────────
-
-const DASHBOARD_ISOLATED_USER_PASSWORD = 'e2e-dashboard-pw-123!';
-
-async function loginAsIsolatedDashboardUser(
-  browser: Browser,
-  adminPage: Page,
-  testPrefix: string,
-): Promise<{ context: BrowserContext; page: Page; userId: string }> {
-  const email = `dash-${testPrefix}-${Date.now()}@e2e-test.local`;
-  const user = await createLocalUserViaApi(adminPage, {
-    email,
-    displayName: 'E2E Dashboard User',
-    password: DASHBOARD_ISOLATED_USER_PASSWORD,
-  });
-
-  const context = await browser.newContext({ storageState: { cookies: [], origins: [] } });
-  const scopedPage = await context.newPage();
-  await scopedPage.request.post('/api/auth/login', {
-    data: { email, password: DASHBOARD_ISOLATED_USER_PASSWORD },
-  });
-  // Force English regardless of the CI browser's default locale — every assertion
-  // in these scenarios matches English card titles (e.g. "Quick Actions").
-  await scopedPage.request.patch('/api/users/me/preferences', {
-    data: { key: 'locale', value: 'en' },
-  });
-
-  return { context, page: scopedPage, userId: user.id };
-}
-
 // ─────────────────────────────────────────────────────────────────────────────
 // Scenario 6: Card dismiss — clicking dismiss hides card; reload keeps it hidden
+//
+// These two tests are the only ones in this file that WRITE
+// `dashboard.hiddenCards`. They previously provisioned a dedicated user inline
+// (PR #1956) to survive the reset hook of a sibling test running in another
+// worker; that is now handled file-wide by `isolatedUserPerWorker` above, so the
+// plain `page` fixture is already the dedicated user's page and the inline helper
+// is gone (Issue #1957).
 // ─────────────────────────────────────────────────────────────────────────────
 
 test.describe('Card dismiss (Scenario 6)', () => {
-  test('Dismissing a card hides it from the dashboard', async ({ page, browser, testPrefix }) => {
-    const {
-      context,
-      page: scopedPage,
-      userId,
-    } = await loginAsIsolatedDashboardUser(browser, page, testPrefix);
-    const dashboardPage = new DashboardPage(scopedPage);
+  test('Dismissing a card hides it from the dashboard', async ({ page }) => {
+    const dashboardPage = new DashboardPage(page);
 
-    await interceptDashboardApis(scopedPage);
+    await interceptDashboardApis(page);
 
     try {
       await dashboardPage.goto();
@@ -572,21 +541,14 @@ test.describe('Card dismiss (Scenario 6)', () => {
       const afterDismiss = dashboardPage.card('Quick Actions');
       await expect(afterDismiss).toHaveCount(0);
     } finally {
-      await uninterceptDashboardApis(scopedPage);
-      await context.close();
-      await deleteUserViaApi(page, userId);
+      await uninterceptDashboardApis(page);
     }
   });
 
-  test('Dismissed card stays hidden after page reload', async ({ page, browser, testPrefix }) => {
-    const {
-      context,
-      page: scopedPage,
-      userId,
-    } = await loginAsIsolatedDashboardUser(browser, page, testPrefix);
-    const dashboardPage = new DashboardPage(scopedPage);
+  test('Dismissed card stays hidden after page reload', async ({ page }) => {
+    const dashboardPage = new DashboardPage(page);
 
-    await interceptDashboardApis(scopedPage);
+    await interceptDashboardApis(page);
 
     try {
       await dashboardPage.goto();
@@ -601,7 +563,7 @@ test.describe('Card dismiss (Scenario 6)', () => {
 
       // Reload the page. Use navigationTimeout (10s) for the heading waitFor since the SPA
       // must fully initialize after a hard reload before the Dashboard heading appears.
-      await scopedPage.reload();
+      await page.reload();
       await dashboardPage.heading.waitFor({ state: 'visible', timeout: 10000 });
 
       // On page load, two contexts fetch preferences independently:
@@ -611,14 +573,12 @@ test.describe('Card dismiss (Scenario 6)', () => {
       //
       // waitForLoadState('networkidle') ensures all pending network requests (including both
       // preference fetches) have completed and React has finished re-rendering before we assert.
-      await scopedPage.waitForLoadState('networkidle', { timeout: 15000 });
+      await page.waitForLoadState('networkidle', { timeout: 15000 });
 
       // The Quick Actions card must be absent — usePreferences applied hiddenCards: ["quick-actions"]
       await expect(dashboardPage.card('Quick Actions')).toHaveCount(0);
     } finally {
-      await uninterceptDashboardApis(scopedPage);
-      await context.close();
-      await deleteUserViaApi(page, userId);
+      await uninterceptDashboardApis(page);
     }
   });
 });
@@ -628,19 +588,10 @@ test.describe('Card dismiss (Scenario 6)', () => {
 // ─────────────────────────────────────────────────────────────────────────────
 
 test.describe('Card re-enable (Scenario 7)', () => {
-  test('Customize button appears when a card is dismissed', async ({
-    page,
-    browser,
-    testPrefix,
-  }) => {
-    const {
-      context,
-      page: scopedPage,
-      userId,
-    } = await loginAsIsolatedDashboardUser(browser, page, testPrefix);
-    const dashboardPage = new DashboardPage(scopedPage);
+  test('Customize button appears when a card is dismissed', async ({ page }) => {
+    const dashboardPage = new DashboardPage(page);
 
-    await interceptDashboardApis(scopedPage);
+    await interceptDashboardApis(page);
 
     try {
       await dashboardPage.goto();
@@ -665,25 +616,14 @@ test.describe('Card re-enable (Scenario 7)', () => {
       // Customize button should now appear
       await expect(dashboardPage.customizeButton).toBeVisible();
     } finally {
-      await uninterceptDashboardApis(scopedPage);
-      await context.close();
-      await deleteUserViaApi(page, userId);
+      await uninterceptDashboardApis(page);
     }
   });
 
-  test('Customize dropdown lists dismissed card and clicking re-enables it', async ({
-    page,
-    browser,
-    testPrefix,
-  }) => {
-    const {
-      context,
-      page: scopedPage,
-      userId,
-    } = await loginAsIsolatedDashboardUser(browser, page, testPrefix);
-    const dashboardPage = new DashboardPage(scopedPage);
+  test('Customize dropdown lists dismissed card and clicking re-enables it', async ({ page }) => {
+    const dashboardPage = new DashboardPage(page);
 
-    await interceptDashboardApis(scopedPage);
+    await interceptDashboardApis(page);
 
     try {
       await dashboardPage.goto();
@@ -712,9 +652,7 @@ test.describe('Card re-enable (Scenario 7)', () => {
       // The button is removed from the DOM entirely when all cards are visible.
       await expect(dashboardPage.customizeButton).not.toBeVisible();
     } finally {
-      await uninterceptDashboardApis(scopedPage);
-      await context.close();
-      await deleteUserViaApi(page, userId);
+      await uninterceptDashboardApis(page);
     }
   });
 });