[OrcaWin]

	Files	Added	Deleted	Net
Test	2	$\color{#1a7f37}{\Huge{\mathbf{+}}}$​113	$\color{#cf222e}{\Huge{\mathbf{−}}}$​3	$\color{#1a7f37}{\Huge{\mathbf{+}}}$​110
Prod	2	$\color{#1a7f37}{\Huge{\mathbf{+}}}$​37	$\color{#cf222e}{\Huge{\mathbf{−}}}$​15	$\color{#1a7f37}{\Huge{\mathbf{+}}}$​22

LoC: +154/-22

Count ADB command time toward the Android boot polling deadline. The loop previously counted only sleeps, so slow census/name/boot-property commands could extend the three-minute wait and an expired successful response could still be accepted. Forward the remaining budget through the existing runner, retain its 60s per-command cap, clip the last sleep and reject late completion. Initial discovery/launch remain outside the polling deadline.

Stacked on #35536 (requested AVD identity). Merge/reconcile that parent first, then retarget/rebase onto main and revalidate; preserve #9950 SDK/network guards.

Validation: 30 configured tests/four suites, strict focused TypeScript and five ordinary hook tasks pass. Exact parent: three deadline regressions fail, timely-boot control passes. Real Node26.8.1 spawn/execFile against owned SDK executable fixtures, seven exact production modules, ten groups per arm at 1/10 boots: 33 adverse calls run 99 polling censuses before vs 33 after; 33 actual stalled subprocesses canceled. Observed polling 2746.78–3066.18ms before vs 303.80–310.60ms after with a 300ms budget; this includes scheduling/termination overhead and is not a strict wall-clock guarantee. Healthy/existing-serial controls preserved; all 715 baseline+committed-final fixture PIDs absent and profiles removed. No physical SDK/device, native Electron UI, macOS/Windows, CPU/RSS claim.

A post-format artifact byte check failed for the earlier pre-hook native run; that evidence is retained as superseded. Native fixed workload was rerun from committed source and rebuilt byte equality verified. Required aggregate queued after #35536; no aggregate pass or Claude takeover claimed. Frozen head 95445f85a699563f2c9c2e95e503b9664651d1a0, parent 8b7d1731bab0fda358defe6fa0619ff937a54dbd. Evidence /home/neil/.local/state/orca-electron-audit/android-avd-boot-deadline/report.md; 51-payload seal c4484bfa26b4154dd015bfc00680e943464885877a3715b1e1e6cefd445c969b.

[OrcaWin]

Thanks. Closing this one. Switching boot polling to a wall-clock deadline fixes an overrun that only shows up when adb calls stall for tens of seconds each, and the existing 60s per-command cap already bounds that. It also creates a user-visible regression: an emulator that finishes booting just after the deadline is reported as a timeout, and the per-command adb timeouts are cut short on the last poll. That doesn't justify around 100 lines of fake-timer tests plus a stack on #35536 and a rebase onto main's localized error copy. The boot identity fix in #35536 will land on its own.

Closing per maintainer triage (low-risk / zero user trade-off / material-gain bar).