ci(smoke): date the main-thread block with a heartbeat

[FerroxLabs]

#1415 established WHICH side hangs on win32-arm64. This dates WHAT blocks it.

Where the investigation stands

readiness stall diagnostics: targets=1 [page:file:///C:/.../wayland-install]
| browser endpoint did NOT answer (CDP command timed out: Browser.getVersion)
- the main process is blocked, not the renderer

The DevTools HTTP endpoint still served /json/list in that same failure, which is served without the main JS thread. So the main process's JS thread is blocked.

What is already excluded: agent detection. It completes in 12278ms and the block outlives it. The app's last log line is kickoff.telemetry about 12s after window creation, then silence for the remaining 600s of the readiness window.

The probe

A blocked JS thread cannot fire its own timer, so the timer is the measurement. A tick that arrives late by more than its own period measures a stall the event loop genuinely suffered. The last tick before silence dates a block that never ends, which is enough to name whatever ran immediately before it.

Gated on WAYLAND_PACKAGE_SMOKE_MARKER, set only by the packaged smoke, so no shipped run pays for it.

Extracted into its own module rather than left inline in index.ts so it can be tested at all. That is the same reason describeDirtyShutdown was extracted.

Three outcomes, all useful

• win-arm64 passes, and v0.13.2 tags immediately
• the heartbeat gaps, and names what ran immediately before the block
• the heartbeat ticks straight through the failure, which falsifies the model - the main thread would not be blocked after all, and the browser endpoint would be unanswered for some other reason

Verification

• full unit suite: 1769 passed, 0 failed
• typecheck 0, scoped oxfmt clean, oxlint 0 warnings
• mutation-proven on BOTH halves: removing the smoke gate fails with "expected [Function] to be null"; disabling stall detection fails because the stall logs as an ordinary tick
• a fourth test pins that ordinary timer jitter is NOT reported as a stall, so the warning means something when it appears

A watchdog that cannot fail would be worse than no watchdog, hence mutating both halves rather than one.

🤖 Generated with Claude Code