[t3dotgg]

Release and CI kept failing on the same handful of tests that pass on rerun. Each one had a real race; this fixes the cause in each instead of raising timeouts. (The release runner split landed separately in #22524. This branch is merged with current main, including #22533.)

Web

• The mouse leaked between tests. Playwright's mouse stays where the last test left it, so a component that mounts under it gets pointerenter and :hover before the test runs. That is the ComposerBannerStack flake (about 14 failures today, [Bug]: Flaky browser and server tests on 4-vCPU CI runners, and a Windows-only update.test.ts failure #22324) and why test(web): park the pointer before RightPanelTabs keyboard tests #21901 had to park the pointer for one test. render() now parks the mouse off the page when its screen disposes, via a resetPointer browser command. It runs on dispose, not before mount: an await before the first mount broke 21 tests that set up state just before render(). The test-only workaround from test(web): fix release browser test failures #22508 is removed.

• Diff workers raced Pierre's 10s startup limit. Each worker loads Shiki on start, about 1.5s idle and over 10s on a loaded runner. PullRequestDetailPanel never needed real workers, so it now uses Pierre's main-thread fallback (first "Add to agent" on one CPU: 7–20s before, 4–6s after), and the 30s timeout from test(web): fix release browser test failures #22508 is gone. StyledDiffCodeView raises the startup limit except in the test of that limit, and its queries stay in the test's own container, because a timed-out test keeps running and later tests clicked its leftover buttons.

• ProjectFaviconPickerDialog. The atom registry drops an unsubscribed atom on its next task. A row asks for its URL while rendering but subscribes after commit, so a URL published in between was lost. The fake asset atoms are now keepAlive. Under CPU load: 2/6 failed before, 10/10 passed after.

• MessagesTimeline Mermaid copy. Only the first Mermaid test failed: it paid Mermaid's cold load (over 10s on the loaded release runner). The file now loads and renders Mermaid once before its tests.

• ProjectSettingsPanel scope. One test drove three menu passes through navigation and ran out of its 15s budget under load. It is now three tests that each mount at their own URL. This drops the check that the menu updates on navigation inside one mount.

Server

• ClaudeAdapterV2 steer hang (all 9 of today's 120s timeouts). A steer waits up to 5s for foreground subagents to move to the background, but that timer starts after the steer reads skill files from disk, so the test's clock move could land before the timer existed. test(server): Claude steer test no longer hangs for 120s under CI load #22533 made the test tick the clock up to 100 times instead, which still fails when the file read outlasts the ticks. The adapter now takes the deadline as the steer's first step, before any I/O, so one 5s clock move always covers it. The test starts the steer at once and awaits a Deferred for the background request instead of polling, and keeps test(server): Claude steer test no longer hangs for 120s under CI load #22533's Deferred for the late answer. With a forced 50ms file-read delay plus 50 scheduler yields, all 6 steer cases still pass.
• transferBudgetV2 30s timeout. Shell streams group and skip unchanged shells, so a split turn may never send an item at the last sequence. The test waits for the final shell instead.
• OpenCodeServerLedger. The fixture read the process before the inner shell had started; it now waits for a ready line. killGroup ignores only ESRCH when the group is already gone.
• AcpAdapterV2 detached PIDs. The mock agent exited before the detached fixture wrote its PID file; it now waits for a published line.

Verification

• Each race was reproduced first: forced delays gave the exact CI errors before each fix and passed after.
• Stress runs on one or two CPUs (taskset) after the fixes: ClaudeAdapterV2 10/10, transferBudgetV2 20/20 (9/12 failed before), AcpAdapterV2 10/10 (8/10 failed before), OpenCodeServerLedger 16/16, diff worker files 3/3 runs, ComposerBannerStack 5/5.
• Full web package: 546 files, 8509 tests passed. Touched server and provider files: 484 tests passed.
• Targeted lint, format, and typecheck are clean.

Made with Claude Opus 5.5 in Claude Code (T3 Code).

🤖 Generated with Claude Code

[macroscopeapp]

Approvability

Verdict: Approved at 917b698

Macroscope's review found this PR approvable — This is primarily a focused CI/test race-remediation change. Its sole production-path edit makes Claude steer timeout accounting bounded from request start, while the remaining edits are isolated test fixtures and browser-test infrastructure.

Notes:

• No code objects were reviewed. Approvability was decided on eligibility alone.

^{You can add or adjust custom eligibility rules. Learn more.}

[github-actions]

Thread transfer impact

✅ Thread transfer remains within every enforced ceiling.

Provider	Metric	Main baseline	This PR	Impact	PR ceiling
Codex	Total thread wire	5.0 KiB	5.0 KiB	−7 B (−0.1%)	6.8 KiB	✅
Codex	Thread snapshot wire	3.8 KiB	3.8 KiB	0 B (0.0%)	4.9 KiB	✅
Codex	Live turn WebSocket wire	1.2 KiB	1.2 KiB	−7 B (−0.6%)	2.0 KiB	✅
Codex	Live turn WebSocket decoded	20.9 KiB	20.9 KiB	0 B (0.0%)	29.3 KiB	✅
Codex	Live turn messages	2	2	0 (0.0%)	8	✅
Claude	Total thread wire	5.0 KiB	5.0 KiB	0 B (0.0%)	6.8 KiB	✅
Claude	Thread snapshot wire	3.8 KiB	3.8 KiB	0 B (0.0%)	4.9 KiB	✅
Claude	Live turn WebSocket wire	1.2 KiB	1.2 KiB	0 B (0.0%)	2.0 KiB	✅
Claude	Live turn WebSocket decoded	21.2 KiB	21.2 KiB	0 B (0.0%)	29.3 KiB	✅
Claude	Live turn messages	2	2	0 (0.0%)	8	✅

Baseline: 5d27ecd · PR result: 917b698 · Source CI: success

Scenario and decoded snapshot size

10 historical turns, 5 command tools per turn, 878.9 KiB retained MCP result per historical turn, and a 1.05 MiB retained result in the measured turn.

• Codex decoded thread snapshot: 108.5 KiB
• Claude decoded thread snapshot: 108.8 KiB

Updated in place by a trusted workflow. PR artifacts are strictly validated and never executed.

[coderabbitai]

No actionable comments were generated in the recent review. 🎉

ℹ️ Recent review info

⚙️ Run configuration

• Configuration used: Path: .coderabbit.config.ts
• Review profile: CHILL
• Plan: Team
• Run ID: e3823f7b-7db4-4978-bbcc-35e18c2cfc59

📥 Commits

Reviewing files that changed from the base of the PR and between 2212a3a and 917b698.

📒 Files selected for processing: 3

• apps/web/src/components/chat/MessagesTimeline.browser.test.tsx
• apps/web/src/components/settings/ProjectFaviconPickerDialog.browser.test.tsx
• apps/web/src/components/settings/ProjectSettingsPanel.browser.test.tsx

Included review availability: This review used your included allowance. Your plan provides up to 10 included reviews per hour; 3 remain after this review.

📝 Walkthrough

This change updates browser test isolation, WebSocket completion checks, process lifecycle synchronization, and Claude steering timing. Claude steering now applies a five-second background-transition deadline measured from the steer request.

Changes: Browser test cleanup

Layer / File(s)	Summary
Reset pointer during browser test cleanup apps/web/test/render.tsx, apps/web/vite.config.ts	The browser test project defines resetPointer, and screen disposal calls it after cleanup.
Scope diff view browser tests apps/web/src/components/diffs/StyledDiffCodeView.browser.test.tsx	Diff view tests use container-scoped queries and configure worker startup limits for test cases.
Update component browser test setup apps/web/src/components/chat/ComposerBannerStack.browser.test.tsx, apps/web/src/components/pullRequest/PullRequestDetailPanel.browser.test.tsx	The banner test checks initial visibility directly. Panel tests mock diff worker construction and remove a custom timeout from the action test.
Stabilize browser test setup apps/web/src/components/chat/MessagesTimeline.browser.test.tsx, apps/web/src/components/settings/ProjectFaviconPickerDialog.browser.test.tsx, apps/web/src/components/settings/ProjectSettingsPanel.browser.test.tsx	Browser tests warm up Mermaid, preserve asset results with keep-alive atoms, and separate project environment scope cases.

Transfer stream completion checks

Layer / File(s)	Summary
Use stream-specific completion conditions apps/server/integration/transferBudgetV2.integration.test.ts	Thread queues use the stored event sequence for completion. Shell queues wait for a matching final projected shell update.

ACP mock process synchronization

Layer / File(s)	Summary
Wait for detached command publication apps/server/scripts/acp-mock-agent.ts	The detached launcher signals when output arrives. The exit path waits for that signal before its delay and process exit.

Claude steering deadline

Layer / File(s)	Summary
Apply the request-start background deadline apps/server/src/orchestration-v2/Adapters/ClaudeAdapterV2.ts, apps/server/src/orchestration-v2/Adapters/ClaudeAdapterV2.test.ts	Steering passes a deadline measured from the steer request to subagent backgrounding. Tests coordinate steering with background-task requests and advance time in five-second steps.

OpenCode process group tests

Layer / File(s)	Summary
Synchronize process group test checks packages/provider-opencode/src/server/OpenCodeServerLedger.test.ts	killGroup attempts group termination directly and ignores ESRCH. The wrapper test waits for output from the inner shell before recording wrapper exit.

Priority: ➖ Normal

Estimated code review effort: 3 (Moderate) | ~25 minutes

Change: Bug fix

Suggested reviewers: juliusmarminge

Merge Risk: ⚪ Minimal · up to 917b6

No concrete merge-blocking risk was identified in the reviewed changes.

Pre-merge checks: 3 · 1

Failed checks: 1 warning

Check name	Status	Explanation	Resolution
Description check		The description thoroughly explains the problems, changes, affected tests, and verification results. It does not provide the required Scope and approval information, including explicit maintainer appr…	Add a Scope and approval section that links the triaged issues or approval discussion and identifies explicit maintainer approval. If this qualifies as an approval exemption, explain why the change is a focused fix and provide the required …

Passed checks: 3 passed

Check name	Status	Explanation
Title check		The title clearly summarizes the primary change: removing timing races that caused recurring CI and release test failures.
Linked Issues check		Check skipped because no linked issues were found for this pull request.
Out of Scope Changes check		Check skipped because no linked issues were found for this pull request.

Full details: Description check

Explanation

The description thoroughly explains the problems, changes, affected tests, and verification results. It does not provide the required Scope and approval information, including explicit maintainer approval or an explanation for why approval was unnecessary.

Resolution

Add a Scope and approval section that links the triaged issues or approval discussion and identifies explicit maintainer approval. If this qualifies as an approval exemption, explain why the change is a focused fix and provide the required scope rationale.

✨ Finishing Touches

📝 Generate docstrings

• Commit to this branch
• Create a new PR

🧪 Generate unit tests (beta)

• Commit to this branch
• Create a new PR

• Autofix · Keep fixing CodeRabbit findings and required CI, and resolving merge conflicts

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

Dismissing prior approval to re-evaluate f0b76d1

Dismissing prior approval to re-evaluate 2212a3a

Dismissing prior approval to re-evaluate 917b698