fix(app): harden view audit completion

[lalalune]

Refs #10196. Follow-up to #10603, which closed the Android WebView full-telemetry gap for this issue.

Summary

• Harden audit:views so the soak now records raw view runtime/render/module-cache telemetry, per-view scorecards, heap samples, frontend/network logs, screenshots, and a Playwright video, then fails on missing paths, first-run blocks, render-guard errors, missing cleanup/evictions, page errors, or unbounded heap.
• Add retained-lazy cacheKey telemetry attribution so evictions/releases can be tied back to registered views and overlay apps.
• Fix the desktop residuals found during final review: /character/documents render/fetch churn, logs hydration/layout-shift attribution, and Settings title/back-button overlap.
• Add focused tests for retained lazy attribution, layout-shift route attribution, and Logs hydration behavior.

Evidence

All artifacts are committed under .github/issue-evidence/10196-views-state/.

• audit:views real app soak: passed 22/22 registered view paths, 44 activations, 33 runtime show events, 32 release evictions, max render count 1, 0 render-guard errors, 0 page errors, heap bounded at ~1.01x.
• audit:views raw artifacts: audit-views-runtime-telemetry.json, audit-views-render-telemetry.json, audit-views-module-cache-telemetry.json, audit-views-heap-series.json, audit-views-navigation.json, audit-views-frontend-log.json, audit-views-network-log.json, audit-views-scorecard.md/json.
• Screenshots/video: audit-views-soak.webm, soak-final.png, representative view-*.png, audit-views-logs-loaded.png, audit-views-settings.png.
• App visual audit: bun run --cwd packages/app audit:app passed 349/349. Touched/reachable views manually reviewed as good; report/contact sheet committed as audit-app-report.json, audit-app-manual-review.md, audit-app-reviewed-contact-sheet.png.
• Live LLM trajectories/audio/native mobile captures: N/A for this follow-up; no prompt, model, action, voice, audio, native-device, or connector behavior changed here. The Android WebView evidence for views/state: the plugin-view "OS" has no mounted-view lifecycle, memory accounting, or render/eviction gate — telemetry rings exist but nothing walks every system/developer/plugin view and asserts on them #10196 was supplied by fix(app): add Android view-runtime soak #10603.

Validation

• bun install --frozen-lockfile passed with no lockfile changes.
• node --check packages/app/scripts/audit-views-soak.mjs passed.
• bunx @biomejs/biome@2.5.1 check on touched source/test files passed.
• Focused UI tests passed: 10 files / 58 tests.
• bun run --cwd packages/ui typecheck passed.
• bun run --cwd packages/app audit:app passed: 349/349.
• bun run verify passed on rebased develop at 2026-07-01T03:24:41Z: 474/474 Turbo tasks plus post-audits and dist-path consumer checks.

[coderabbitai]

Important

Review skipped

Auto reviews are disabled on this repository. Please check the settings in the CodeRabbit UI or the .coderabbit.yaml file in this repository. To trigger a single review, invoke the @coderabbitai review command.

⚙️ Run configuration

Configuration used: Organization UI

Review profile: CHILL

Plan: Pro

Run ID: 6bfb3bd7-0a37-4321-94c3-485978f98e4b

You can disable this status message by setting the reviews.review_status to false in the CodeRabbit configuration file.

Use the checkbox below for a quick retry:

• 🔍 Trigger review

✨ Finishing Touches

🧪 Generate unit tests (beta)

• Create PR with unit tests
• Commit unit tests in branch fix/10196-view-audit-completion

---

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

• X
• Mastodon
• Reddit
• LinkedIn

_{Comment @coderabbitai help to get the list of available commands.}

[greptile-apps]

Your trial has ended. Reactivate Greptile to resume code reviews.

[NubsCarson]

Independent validation for #10651 on head cf86c0b582:

• git diff --check origin/develop...HEAD && git diff --check passed.
• node --check packages/app/scripts/audit-views-soak.mjs passed.
• bun install --frozen-lockfile --ignore-scripts passed with no lockfile changes.
• bunx @biomejs/biome check on the 14 touched source/test files passed.
• bun run build:core passed: 64/64 tasks.
• bun run --cwd packages/ui test -- src/components/pages/LogsView.test.tsx src/hooks/useLayoutShiftMonitor.test.ts src/retained-lazy.churn.test.ts src/retained-lazy.test.tsx passed: 4 files, 13 tests.
• bun run --cwd packages/ui typecheck passed.

I did not rerun the full committed audit:views/audit:app visual artifact capture in this pass; the PR branch already includes those artifacts and the focused code gates above did not expose a regression. GitHub checks are still pending/skipped/pass only from the queue snapshot.

[lalalune]

Focused review/validation from my local pass on current head cf86c0b58282:

bun run --cwd packages/ui test -- src/components/pages/LogsView.test.tsx src/hooks/useLayoutShiftMonitor.test.ts src/retained-lazy.test.tsx src/retained-lazy.churn.test.ts
bunx @biomejs/biome check packages/app/scripts/audit-views-soak.mjs packages/ui/src/App.tsx packages/ui/src/components/apps/AppWindowRenderer.helpers.ts packages/ui/src/components/character/CharacterHubView.tsx packages/ui/src/components/pages/DocumentsView.tsx packages/ui/src/components/pages/LogsView.tsx packages/ui/src/components/pages/LogsView.test.tsx packages/ui/src/components/pages/SettingsView.tsx packages/ui/src/components/views/ViewTelemetryProfiler.tsx packages/ui/src/hooks/useLayoutShiftMonitor.ts packages/ui/src/hooks/useLayoutShiftMonitor.test.ts packages/ui/src/retained-lazy.tsx packages/ui/src/retained-lazy.test.tsx packages/ui/src/retained-lazy.churn.test.ts
bun run --cwd packages/ui typecheck
bun run --cwd packages/app typecheck

Results: focused UI tests passed (4 files, 13 tests); Biome pass; packages/ui typecheck pass; packages/app typecheck pass. I also manually opened the committed contact sheet plus audit-views-logs-loaded.png, audit-views-settings.png, and soak-final.png; the touched screens are visible and I did not see overlap/clipping in those captures.

Evidence caveat before using this as final #10196 closeout proof: .github/issue-evidence/10196-views-state/audit-views-frontend-log.json contains repeated browser console Failed to load resource 404s, for example /api/connectors/google/accounts, /api/transcripts, and /api/database/status; audit-views-network-log.json also contains multiple net::ERR_ABORTED requests. Those may be expected optional-service/nav-churn noise, but the evidence should either classify them as expected or the audit should fail/filter them explicitly. I would not call the evidence fully clean while those entries are unexplained.

[greptile-apps]

Your trial has ended. Reactivate Greptile to resume code reviews.

[lalalune]

Pushed f6abc6d90a24 to address the evidence caveat I noted above.

What changed:

• audit-views-soak.mjs now classifies network entries and writes audit-views-network-summary.json.
• Future audit:views runs fail on any unclassified network failure.
• The two expected classes are explicit: fast route-churn net::ERR_ABORTED requests and optional local-service 404s for /api/connectors/google/accounts, /api/transcripts, and /api/database/status.
• The committed evidence now includes the summary: 59 total network entries, 36 expected navigation aborts, 23 expected optional-route 404s, 0 unexpected. The scorecard and soak JSON reference the new artifact.

Validation:

• node --check packages/app/scripts/audit-views-soak.mjs
• bunx @biomejs/biome check packages/app/scripts/audit-views-soak.mjs
• jq empty .github/issue-evidence/10196-views-state/audit-views-network-summary.json .github/issue-evidence/10196-views-state/audit-views-soak.json
• git diff --check origin/develop...HEAD

I did not rerun the full visual/audit capture after this patch; this is script hardening plus classification of the already committed raw network log.

[NubsCarson]

Independent latest-head validation for #10651 on f6abc6d90a2489d719bc52873e57b1b9c730afa0.

This reruns validation after the follow-up that classifies network noise in the view audit evidence. I used a fresh worktree; after bun install --frozen-lockfile --ignore-scripts, I had to run the normal generated-output prep (bun run build:core, then bun run --cwd packages/shared build:i18n) before UI/app tests, because the fresh worktree initially lacked packages/shared/src/i18n/generated/validation-keyword-data.js.

Latest-head checks green:

• git diff --check origin/develop...HEAD
• node --check packages/app/scripts/audit-views-soak.mjs
• jq empty .github/issue-evidence/10196-views-state/audit-views-network-summary.json .github/issue-evidence/10196-views-state/audit-views-soak.json
• bunx @biomejs/biome check packages/app/scripts/audit-views-soak.mjs packages/ui/src/App.tsx packages/ui/src/components/apps/AppWindowRenderer.helpers.ts packages/ui/src/components/character/CharacterHubView.tsx packages/ui/src/components/pages/DocumentsView.tsx packages/ui/src/components/pages/LogsView.tsx packages/ui/src/components/pages/LogsView.test.tsx packages/ui/src/components/pages/SettingsView.tsx packages/ui/src/components/views/ViewTelemetryProfiler.tsx packages/ui/src/hooks/useLayoutShiftMonitor.ts packages/ui/src/hooks/useLayoutShiftMonitor.test.ts packages/ui/src/retained-lazy.tsx packages/ui/src/retained-lazy.test.tsx packages/ui/src/retained-lazy.churn.test.ts
• bun run build:core — 64/64 successful
• bun run --cwd packages/ui test -- src/components/pages/LogsView.test.tsx src/hooks/useLayoutShiftMonitor.test.ts src/retained-lazy.churn.test.ts src/retained-lazy.test.tsx — 4 files / 13 tests passed
• bun run --cwd packages/ui typecheck
• bun run --cwd packages/app typecheck

Worktree is clean. GitHub checks remain queued/UNSTABLE; I do not see a local blocker in this pass.