fix(platform): expose pending-question read failures

[yannickmonney]

A failed pending-question read no longer presents a live automation task as ordinary Working. The panel shows an accessible error and Try again, keeps that recovery surface during retry, and preserves cached question cards and answer drafts. A successful empty read restores Working; failed-read memory stays scoped to the run/task. Copy ships in EN/DE/FR.

Verification:

• 72 focused component tests passed with one worker: subject panel and task-modal draft/repeat/limits consumers.
• Scoped TypeScript, type-aware oxlint, oxfmt, diff checks and manual-reference lint passed.
• Chromium checked EN/DE/FR mobile layout and keyboard retry, busy state and recovery against the actual panel with mocked query/mutation boundaries. Visual aspect analyzer: 100/100, no defects.
• Manual register insertion uses the untouched “Task attachment uploads overlap…” anchor. Inspected open PR diffs and compared merge-tree results for 43 branch heads against main: no new register conflicts (some old branches already conflict with main).

Closes #3850

Current-main rebase

Replayed the previously accepted source 8d4b1ed362a24cc520464e07bf5320e1e2167a6a onto main d1373d84cd56972501403f62145ec52e6f65d44a, including the merged shared CI repair in #4625. The accepted feature payload and all current-main changes are preserved in one atomic commit. Configured commit and conflict checks pass; earlier behavioral proof remains recorded above. All seven native required checks and full merge-group validation remain required for this new source.

Maintenance replay: preserves the accepted feature payload on current main 7d178ca. Includes the merged #4649 Knip cleanup and the exact independently accepted one-line shared CLI inventory repair from #4654 (252f0df), which is still pending native merge on main. The fixed suite inventory keeps its discovery and source/compiled phase guards. Existing feature proof is retained; no fresh full-feature/full-workspace test or hosted-green claim. Native required checks remain mandatory.

[yannickmonney]

TALE-282 / #3850 implementation report, with TALE-359 (3729d02a-eb44-42b2-a68e-1bec2cebbcd6) fleet-report fallback:

Current-main repair is ready for independent review in this PR. Pending-question read failures expose an accessible error and Try again instead of Working; retry, cached questions/drafts, successful empty/question recovery, and run/task isolation are covered. EN/DE/FR copy is included.

Regression replay against the unchanged main panel fails the unavailable-question and cached-question read-failure cases (2 expected failures); the repaired panel passes.

Local verification: 72 focused component/consumer tests + 24 locale tests passed; scoped TypeScript, type-aware oxlint, oxfmt, manual-reference lint and diff checks passed. Real Chromium checked mobile EN/DE/FR plus keyboard retry/busy/recovery. The visual-aspect-analyzer supplied the browser visual gate: 100/100, no defects. This used the actual component in an isolated fixture, not a deployed production system.

Documentation insertion anchor: “Task attachment uploads overlap while their drain, read and replace steps are”. Inspected open-PR diffs and compared git merge-tree results against main for 43 branch heads: no new documentation-register conflicts. Older branches' pre-existing main conflicts are recorded in the attached delivery evidence.

CI checks were watched and remain queued: 21 native jobs with no assigned runners, plus 7 expected skipped candidate jobs. No completed native verdict or failure log is available; CI is not verified green. Known main lint breakage has a separate repair in #4386, but no baseline-failure claim is made for this PR without its logs. No self-acceptance or merge.

The organization task-tool bridge returned not_granted for workspace_status, so this comment is the requested reporting fallback for both task IDs. Delivery evidence is saved under the exact TALE-282 delivery directory.

[yannickmonney]

Independent exact-head review of f54b5a3: PASS. The focused platform UI suite passes 37/37 with one worker. The added coverage verifies settled pending-question read failure renders an accessible error and Try again, suppresses ordinary Working, keeps cached question/draft through retry, recovers to both empty and answerable success, and scopes failure state across runs/tasks. EN/DE/FR keys are present; the manual register insertion uses a distinct existing anchor. Existing CI results are green for Performance, Opengrep/Opengrep OSS, Build/E2E scope PR checks; the remaining Checks jobs were still pending when reviewed. No findings.