[yaacovcorcos]

What Changed

Test-only. Five browser tests that fail intermittently on busy CI runners now wait for what they assert.

• updateRestartFlow "closing the popover does not swallow a later keyboard focus" (Scient). Moving the pointer off the update button starts its 150 ms hover-close timer, and opening the popover from keyboard focus does not cancel that timer. When Tab lands inside the window, the timer closes the popover the focus just opened. The test now waits out the delay before pressing Tab, as a person would. Reproduced: with the delay temporarily stretched to 2 s (not committed), the old test fails at the CI line and the new one passes.
• WorkflowOpeningBlock "shows a short script as highlighted code" (upstream). Upstream's own fix, pingdotgg/t3code#21699, cherry-picked unchanged with -x and its author kept, so the next alignment merges cleanly.
• PullRequestSummaryTab filter test (upstream). It looked up the filter menu's items synchronously right after the click that opens the menu. It now waits for them, using vi.waitFor.
• RightPanelTabs keyboard "uses a custom binding from a text field and closes on Escape" (upstream). It still steps fake animation frames, but now polls in real time too, for up to 5 s instead of 10 instant frames. I did not confirm the root cause. This only lengthens the wait and cannot make the test stricter.
• AttachmentFilePreview "keeps a renewed URL" (upstream). The source view is lazy-loaded, so the wait for it is now 5 s instead of the default 1 s. This is likely the cause, but it is not reproduced.

Each change to an upstream test carries a SCIENT-FORK: comment.

Why

The upstream alignment (#593) added many browser tests. Since it merged, at least one of these five has failed in nearly every merge-queue and PR run today, on PRs that don't touch them. Each failure drops a PR from the queue: #599 and #604 were both dropped this way. The failures are:

Test	Seen failing on
updateRestartFlow	#596 PR CI; #593, #599 queue runs; #601 PR CI
WorkflowOpeningBlock	#604 queue run; #603 PR CI
RightPanelTabs keyboard	#599, #604 queue runs; #603 PR CI
PullRequestSummaryTab	#593 queue run
AttachmentFilePreview	#601 PR CI

Not covered here:

• WordFileExport (server) failed once on test(server): keep server tests out of the source checkout #601, reporting "A file in the LaTeX project changed while exporting".
• The Windows packaging voice qualification timed out once on fix(antigravity): no console window on Windows; say when the runtime check runs out of time #603.

Documentation Impact

Documentation impact: None — test-only timing changes; no behavior or workflow change.

Verification

On Windows 11 with headless Chromium:

• The five changed test files: 31 tests passed.
• The 2 s close-delay reproduction: the old updateRestartFlow test fails, and the fixed one passes.
• Web typecheck: clean, with suggestions only.
• vp fmt --check and vp lint on the changed files: clean.

Checklist

• This PR is small and focused
• I explained what changed and why
• I included before/after screenshots for any UI changes (no UI change)
• I included a video for animation/interaction changes (none)

🤖 Generated with Claude Code

[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	5.6 KiB	5.6 KiB	0 B (0.0%)	6.8 KiB	✅
Codex	Thread snapshot wire	4.3 KiB	4.3 KiB	0 B (0.0%)	4.9 KiB	✅
Codex	Live turn WebSocket wire	1.3 KiB	1.3 KiB	0 B (0.0%)	2.0 KiB	✅
Codex	Live turn WebSocket decoded	26.2 KiB	26.2 KiB	0 B (0.0%)	29.3 KiB	✅
Codex	Live turn messages	2	2	0 (0.0%)	8	✅
Claude	Total thread wire	5.6 KiB	5.6 KiB	0 B (0.0%)	6.8 KiB	✅
Claude	Thread snapshot wire	4.3 KiB	4.3 KiB	0 B (0.0%)	4.9 KiB	✅
Claude	Live turn WebSocket wire	1.3 KiB	1.3 KiB	0 B (0.0%)	2.0 KiB	✅
Claude	Live turn WebSocket decoded	26.6 KiB	26.6 KiB	0 B (0.0%)	29.3 KiB	✅
Claude	Live turn messages	2	2	0 (0.0%)	8	✅

Baseline: 129f4af · PR result: 221544d · 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.
Full compact HTTP snapshot with live cursor.

• Codex decoded thread snapshot: 133.1 KiB
• Claude decoded thread snapshot: 133.4 KiB

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