fix(platform): retry failed website page loads without skipping

[yannickmonney]

A failed Website details Load more request previously advanced the offset before receiving its page, so the next click skipped those results. The next read now starts at the number of accepted rows; refresh bookkeeping advances only after a response is accepted. Loaded rows remain visible, and a later-page failure exposes a live error and Try again, using existing EN/DE/FR messages.

The regression covers repeated offset-20 failures, retry and recovery, subsequent offset-40 paging, accessibility, and exactly one copy of each row. The automation register records this coverage in its reserved paragraph.

Validation: 65 targeted UI tests passed across the dialog, row actions and websites table; 75 failure-reporting guard tests passed. Scoped oxlint with --type-aware --type-check, oxfmt, git diff --check, and the manual register lint passed. Browser evidence and CI status follow in a comment.

Closes #3851

Current-main rebase

Replayed the previously accepted source bc594df9386474e33d13f897e422a41b308fa782 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-283 / TALE-359 dispatch update: PR #4377, head 725fb014532a46d4df20d98216c3b0bfef1d21c1.

Failed Load more requests retain accepted rows and request the same offset on retry. The dialog exposes a live error with Try again using existing EN/DE/FR text. The regression covers repeated failure, retry, recovery, continued paging and no duplicates; the reserved automation register paragraph is added.

Local verification: 65 targeted UI tests and 75 failure-reporting guards passed; scoped type-aware oxlint/type checking, formatting, manual register lint and diff checks passed. CI is running with no failures currently reported. Browser verification confirms offset 20 is retried and recovers; the final visual analyzer assessment is still being checked.

Workspace task reporting was refused: workspace_status returned unavailable / not_granted. This PR comment supplies the requested TALE-283 and TALE-359 update. Independent review remains pending; no acceptance or merge performed.

[yannickmonney]

TALE-283 / TALE-359 final handoff: #4377 at head 725fb014532a46d4df20d98216c3b0bfef1d21c1.

The fix retains accepted rows and retries a failed later page at the same offset, with a persistent accessible error and Try again. Existing EN/DE/FR messages are reused. The regression and reserved automation register paragraph are included.

Verification: 65 targeted UI tests, 75 failure-reporting guard tests, scoped oxlint --type-aware --type-check, oxfmt, manual register lint and diff checks passed. Chromium confirmed offsets [0, 20, 20, 40] and 60 rows after recovery.

CI's code checks passed. Build db failed before any source build, during Setup Docker builder: Docker Hub timed out fetching moby/buildkit:buildx-stable-1. Consequently CI ready (Build) also fails. The failed-job rerun was refused: GitHub's rerun-failed-jobs endpoint returned HTTP 403, Resource not accessible by personal access token. A reviewer with Actions write access needs to rerun the failed jobs in https://github.com/tale-project/tale/actions/runs/37344428588. No infrastructure or credential changes were made.

The visual-aspect-analyzer gate remains open: both the successful control dialog and later-page failure state score 75 with visibility/flicker diagnostics in the dialog title/status regions. The screenshot confirms the visible error and retry layout; this is not a claimed score-100 visual pass. Evidence, source snapshots, patch, logs and reports are in this task's delivery box.

Workspace task tools were refused (workspace_status: unavailable/not_granted), so this comment reports the result for TALE-283 and TALE-359. Independent review remains pending; no acceptance or merge performed.

[yannickmonney]

ACCEPT — independent review of exact head 725fb014532a46d4df20d98216c3b0bfef1d21c1 for PR #4377 / TALE-283 (closes #3851).

No blocking source findings. Failed later-page reads retain accepted rows. Retry derives its offset from accepted rows rather than advancing before success; the regression exercises offsets [0, 20, 20, 20, 40], repeated failures, recovery, resumed paging, and exactly one copy of all 60 rows. Accepted responses alone clear the error and update refresh bookkeeping; the existing overtaken-refresh regression remains green. The live Alert stays mounted while retrying, and the shared Button disables activation and exposes busy state. Existing EN/DE/FR toast.fetchPagesError and common.actions.tryAgain translations cover both messages.

Composition rechecked against main abb19b9d82c5a73ca6163db53773f09f3a37ba4f: clean merge-tree, no component collision. The automation paragraph remains immediately after “Website content search clears…” and before Inbox discovery.

Verification with existing dependencies:

• /opt/node/bin/node Vitest, one worker: all 65 tests pass across website-view-dialog, website-row-actions and websites-table. Used the repository UI config with a review-only filesystem allowance for shared dependencies and --testTimeout 30000. The first completed run passed 64/65; the new regression hit the default 5000ms timeout. The bounded rerun passed all assertions.
• Scoped oxlint --type-aware --type-check --threads 1 on both changed TSX files: exit 0, no diagnostics.
• Scoped oxfmt: both TSX files pass. Markdown is excluded by repository configuration. git diff --check passes.

CI is settled, but not entirely green: Build db failed during Setup Docker builder because connecting to https://registry-1.docker.io/v2/ timed out; source compilation never started in that job. CI ready (Build) failed consequentially. Source lint, format, type, unit/UI, platform build and E2E checks passed. Infrastructure failure evidence: https://github.com/tale-project/tale/actions/runs/37344428588/job/111879783684 . ACCEPT is the source-review verdict, not a claim that the CI merge gate is green.

Limits: no live browser/screen-reader verification in this light review. The component mock fixes isPending to false, so retry-time error persistence and busy-button behavior were verified by implementation inspection rather than that mock.

TALE-283 and TALE-359 (3729d02a-eb44-42b2-a68e-1bec2cebbcd6) delivery fallback: workspace_status returned unavailable / not_granted, so direct task comments could not be posted. This PR comment carries the single verdict for both destinations as requested. No code changes, push, merge, CI rerun or status change.

[yannickmonney]

TALE-896 delegated merge report for TALE-283; fallback report for TALE-359 and TALE-360 because workspace_status returned not_granted.

SKIPPED #4377 at requested exact head 725fb01. It remains OPEN, but mergeStateStatus is UNSTABLE, not CLEAN. Two checks failed: Build db and CI ready (Build), run https://github.com/tale-project/tale/actions/runs/37344428588. Both logs were read. Build db failed contacting https://registry-1.docker.io/v2/: request canceled while waiting for connection / Client.Timeout exceeded while awaiting headers. CI ready (Build) failed with: build expected success got failure; smoke-test and image-validate expected success got skipped; vulnerability-scan expected completed advisory execution got skipped.

This is a registry connection timeout and its dependent readiness failure, not proven fallout from main 1994b8a or use-compact-age.test.ts. The permitted exception does not apply. Stopped at the eligibility/CI gates; no composed-tree tests or merge attempted, and no rerun, cancellation, approval or status change.

Slice result: #4371 merged as 1a86f23 (11 targeted tests passed; #3882 CLOSED), #4378 as 966fd6a (126 passed; #3864 CLOSED), #4379 as 9e8628a (14 passed; #3868 CLOSED). Each merged PR passed exact-head/CLEAN, terminal checks, seven CI ready contexts, uncached platform Unit/UI shards, clean merge-tree and duplicate-free EN/DE/FR catalog gates. Detailed reports are on each PR. #4377 remains open for a later authorized pass.

[yannickmonney]

TALE-904 delegated merge report; fallback for TALE-283, TALE-359 (3729d02a-eb44-42b2-a68e-1bec2cebbcd6), and TALE-360 (c42a085c-35c6-459f-8006-bdae1a0dc5e1): workspace_status returned unavailable/not_granted, so task comments cannot be posted through the granted service.

Skipped #4377. It remains OPEN at accepted head 725fb014532a46d4df20d98216c3b0bfef1d21c1. Read both failing job logs. Build db job 111879783684 failed in Setup Docker builder: Docker Hub registry request timed out before Login to GHCR or Build and push. CI ready (Build) job 111882447702 failed because build failed and downstream image/smoke/scan jobs skipped. This is not main’s use-compact-age lint breakage, so the card’s condition authorizing update-branch is not satisfied. No branch update, rerun, approval, status change, or merge performed. Issue #3851 remains open.

Evidence: https://github.com/tale-project/tale/actions/runs/37344428588/job/111879783684 and https://github.com/tale-project/tale/actions/runs/37344428588/job/111882447702. Full logs are in the TALE-904 delivery box.