feat(vault): keep refund transaction links after confirmation

[jonybur]

Outcome

Inactive Vaults keeps refund transaction hashes visible before and after confirmation. The refund success dialog also shows the hash and a copy control. The branch includes current main. Fresh CI and Jerome's approval are pending.

Change

• Keep completed refunds in Inactive Vaults.
• Cache confirmed refund hashes and exclude confirmed refunds from further polling.
• Keep pending broadcast hashes available for display.
• Resolve conflicts with the current phone and vault recovery changes.
• Make the refund review test wait for the price to load before checking it.

Safety

The hash is for display. Existing action checks block a second withdrawal. Invalid hashes and known PegIn hashes are omitted. Old timestamp-only cache entries remain valid. Cache expiry stays unchanged. An in-session reorganization does not restart polling. Current phone link rules and the open redeem success dialog remain supported.

Verification

• Full vault suite: 5,096 passed, two skipped. After the test timing fix, all 11 refund modal tests passed.
• Six refund screen checks passed across desktop, 390 px, and 360 px widths.
• Vault lint and type checks passed. Lint reports 138 warnings and no errors.
• All three PR updates merge together without conflicts. The combined vault suite passed 5,102 tests, with two skipped.
• A live wallet refund remains untested. Fresh GitHub CI must pass before merge.

Links

• Refund visual test
• Separate sidebar change: fix(vault): show pending deposit progress in the sidebar #2671.
• Shared workflow fixes: ci: waive the deliberate pull_request_target in pre-review-check #2668 and ci: fix zizmor and ci-doctor findings #2669.

The following record is from the earlier review. The checks above cover this update.

Pre-review: 1 finding · 1 fixed · 0 open (0 merge-blockers) · 0 undecided · lint/test passed

Full review covered all 21 changed files. Uncached affected lint and tests passed. No source file was rewritten by the checks.

#	Finding	Severity	Outcome
1	Unused explorer button label	normal	fixed

[github-actions]

🔐 Commit Signature Verification

✅ All 1 commit(s) passed verification

Commit	Author	Signature	Key Type	Key Check
d7906f7a5ab2	Jonathan Bursztyn	✅	sk-ssh-ed25519	✅

Summary

• Commits verified: 1
• Signature check: ✅ All passed
• Key type enforcement: ✅ All sk-ssh-ed25519

---

Required key type: sk-ssh-ed25519 (FIDO2 hardware key)

Last verified: 2026-10-08 23:43 UTC

[greptile-apps]

Confidence Score: 4/5

[Medium risk] Adds refund transaction tracking to the deposit activity feed.

The PR appears safe to merge with non-blocking corrections to refund-row timing, hash-path coverage, and stale-link handling.

Findings

1.  Refunds appear under deposit dates ▶
2.  Visual test bypasses hash selection ▶
3.  Cached links can become stale ▶

Summary

The PR carries refund transaction hashes from broadcast and HTLC polling into Activity, retains confirmed hashes across reloads, and creates a refund row when indexed history lacks a deposit row. It also adds desktop and mobile visual captures.

• Synthetic rows should use refund timing rather than deposit timing.
• The new hash paths need direct tests, and confirmed links need in-session revalidation after a reorganization.

Diagram

%%{init: {'theme': 'neutral'}}%%
flowchart LR
  Broadcast[Refund broadcast] --> Pending[Address-scoped pending record]
  Outspend[HTLC outspend polling] --> Result[Deposit polling result]
  Pending --> Result
  Outspend --> Cache[Confirmed refund cache]
  Cache --> Result
  Result --> Activity[Indexed or synthetic Activity row]
  Activity --> Mempool[Mempool transaction link]

Loading

_{Reviews (1) · Last reviewed commit: "feat(vault): link refund transactions to..."}

[github-actions]

Visual regression

🖼️ 6 of 447 screens render differently than the merge-base.

vault - 6 of 117 screens changed

vaults-refunded - 3 screens, 3 size-changed

desktop - 1280x800 to 1280x826 - before on top, after underneath

mobile-360 - 360x1014 to 360x1280 - before on the left, after on the right

_{1 more changed screen in this group - listed in full below.}

vaults-refunding - 3 screens, 3 size-changed

desktop - 1280x800 to 1280x826 - before on top, after underneath

mobile-360 - 360x1014 to 360x1280 - before on the left, after on the right

_{1 more changed screen in this group - listed in full below.}

All 6 changed screens

### vault
6 of 117 screens changed:
  size-changed  vaults-refunded--desktop.png
  size-changed  vaults-refunded--mobile-360.png
  size-changed  vaults-refunded--mobile.png
  size-changed  vaults-refunding--desktop.png
  size-changed  vaults-refunding--mobile-360.png
  size-changed  vaults-refunding--mobile.png
### storybook
No visual changes across 330 screens.

_{4 screens pictured, cropped to the part that changed, published at most 960px wide so the set fits in a comment - a sample across the worst-hit groups on each surface, not a ranking. Full-resolution before / after / diff for every screen: open the run, download visual-report, open index.html.}

_{If these changes are intentional there is nothing to update - this repo stores no baseline images. The baseline is recomputed from the merge-base on every run.}

_{Compared d7906f7a5ab2061d13d64c897fd8ff8cbfdafe8d against the merge-base. Images are pruned when this pull request closes.}

[jrwbabylonlab]

Change looks good. One thing: this keeps every refunded vault in the mempool poll forever. Can we store the refund txid in the refunded cache instead and keep excluding confirmed ones from the poll?