[Frontend][Inbox] Surface batch triage polling expiry

[Chris0Jeky]

Summary

• report when Legacy Inbox batch-triage checking reaches its 60-second deadline with unresolved captures
• keep the receipt truthful: the accepted triage may still be running, and Refresh Detail is the recovery action
• render expiry as a persistent warning, not a failed action
• leave terminal completion and explicit polling cleanup quiet

This closes the remaining #2230 acceptance without extending client polling or relabelling an accepted triage as failed.

Current-base proof

• exact head: 390a311a29c1d2c0d092a821f1f6b502bce57335
• reviewed base: 98f3fbd145dd6c48e3b47da56fc110d3e975b4c1
• 126/126 Inbox/API/orchestrator tests passed
• 60/60 toast/render tests passed
• scoped ESLint passed
• explicit typecheck and production build passed
• diff hygiene passed
• fresh current-base interaction review found no CRITICAL/HIGH blocker
• automatic Codex review completed on this head with no new finding
• Smart CI run 33781168216, CI Extended run 33781131953, and Required CI run 33781131973 passed; the first E2E attempt stopped before tests when two Microsoft Ubuntu feeds returned 403, and the same-head failed-job rerun passed E2E Smoke

Non-blocking residuals:

• concurrent polls can each publish one truthful warning, and an older poll can replace the shared batchError; the visible receipt stays truthful and batchError has no production template consumer, so per-batch error ownership is declined as out of scope
• batch selection clears after submission, so the user may need to reopen a capture before Refresh Detail is visible

No browser, live-provider timing, or screen-reader run was performed locally. Deterministic fake-timer regressions exercise the owned deadline contract, and hosted E2E Smoke passed.

Closes #2230

[chatgpt-codex-connector]

Codex Review Summary

This comment shows the latest Codex review activity on this pull request.

Review	Status	Commit	Review trigger
📝 Code Review	✅ Completed 2026-09-03T16:53:48.527329Z	390a311	New commits

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review" or "@codex security review".

Codex reacts with 👀 while any review is running, comments if it has suggestions, and reacts with 👍 once all reviews finish with no findings.

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: 80e4811372

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

[Chris0Jeky]

Independent exact-diff review at 80e481137253c843d684a70e61e8651ea6f71a5f: no CRITICAL/HIGH finding.

Verified independently: capture store 57/57, Inbox orchestrator 55/55, typecheck, and diff check. toast.error defaults to duration 0, so this is a persistent receipt.

Nonblocking review disposition:

• declined operation-scoped timeout copy in this slice: concurrent poll A can outlive completed poll B, but the generic receipt remains truthful and does not identify or mutate the wrong capture
• declined a second deadline-specific AbortSignal test: the deadline calls the same stop() path already covered by the existing in-flight abort regression; the new deadline regression proves that path is selected and emits exactly one receipt

Hosted exact-head CI is now the remaining gate.

[Chris0Jeky]

Fix-diff independent review at 3ec891eea92ecb7b3a08deb859ca0227e85b36dd: no CRITICAL/HIGH finding.

The reviewer verified that toast.warning(message, 0) is persistent in the shared store, renders as a non-failure in both Legacy and Paper, has no assertive error semantics, and shows no countdown. Focused capture tests passed 57/57; toast-store and both renderer suites passed 60/60; diff check passed.

The reviewer noted that the terminal-completion regression does not separately assert the warning mock is quiet. Classified nonblocking and declined at the fix-review ceiling: the test completes before the deadline and advances beyond it, while source inspection confirms completion clears that same deadline timer. The explicit-stop regression does assert both error and warning remain absent.

[Chris0Jeky]

Fresh current-base interaction review completed at 390a311a29c1d2c0d092a821f1f6b502bce57335 against 98f3fbd145dd6c48e3b47da56fc110d3e975b4c1.

No CRITICAL/HIGH blocker was found. The deadline path checks completion before publishing one persistent warning and batchError, then aborts and stops polling. Terminal completion and explicit stop stay quiet, and the current request guards prevent late list or detail writes.

Current-base proof passed: 126/126 Inbox/API/orchestrator tests, 60/60 toast/render tests, scoped ESLint, explicit typecheck, production build, and diff hygiene.

Non-blocking review note: concurrent polls share batchError, so an older poll can replace newer shared text. The warning remains truthful for the older accepted work and batchError has no production template consumer. Classified P2 and declined as outside #2230's expiry-disclosure acceptance. Hosted exact-head CI and automatic review remain pending.

[Chris0Jeky]

Coordination note on the parked state. Evidence only; the restart decision stays with the owning lane.

This head has no file-level interaction with anything that landed under it

Merge base is a127dee4f; main has advanced 27 commits since. Computed locally, the intersection of the files this PR changes (src/store/captureStore.ts and its spec) with the files main changed over those 27 commits is empty. The third 2026-09-03 block was two MCP backend slices, a board-refresh arbitration slice and an E2E path fix, none of which touch the capture store. Commit distance here is not evidence of staleness in the change itself.

The only red is #2401, not this branch

Smart CI / Required Gate failed on run 33754458696 with:

• base-sha-mismatch — plan e1ea14fad vs event 2647c3af0
• trust-mismatch — plan says T3, the event re-derives T1

The plan's base is an ancestor of the gate event's base, so main moved between the Plan job and the Gate job of the same run. The identical two-line verdict appeared on PR #2408 (run 33736889079) the same day with unrelated branch content. Both are recorded on #2401.

Smart CI / Required Gate is not an enforced branch-protection context. Measured today, main requires exactly three, all security. The verdict here is shadow-mode observation evidence, and it does count against SC-4's clean-window condition, which is why #2401 now sits first on the release chain.

Note for the #2230 restart contract

The reviewed acceptance (a truthful timeout receipt at the 60-second deadline that does not relabel triage as failed) is unaffected by everything that landed. A current-base run plus the usual disposition should be sufficient; no re-derivation of the polling behaviour is indicated by anything on main.

[Chris0Jeky]

Required CI run 33781131973 reached E2E only after all 16 prerequisite jobs passed. E2E did not start the Taskdeck test suite: npx playwright install-deps chromium failed during apt metadata refresh because both packages.microsoft.com Noble feeds returned HTTP 403. The checked-out merge SHA was f9ffdd0554ae03e5d92cfae9db7a3a369cab422c, combining head 390a311a with base 98f3fbd14.

This is an external runner dependency-install failure, not a product-test result. I am rerunning the failed job once on the same exact head; merge remains blocked until that rerun is green.

[Chris0Jeky]

Required CI run 33781131973 attempt 2 passed on the unchanged exact head. The rerun cleared Playwright dependency installation and completed E2E Smoke successfully. No product test failed in either attempt.