Fixes the Windows deferred bash wait test so it waits for the deferred wait to actually enter instead of racing shell startup deadlines.
Replaces the 10-second startup budget and separate 3-second startup check with a single 120-second hang cap, applied only in this test; no product timeouts change.
Panics if the bash command completes before the wait enters, and rejects tasks already terminal before lock contention, so the held-lock observation is always exercised.
Adds a benchmark descriptor for the land-deferred-wait-causal slice.
Confidence Score: 5/5
The PR appears safe to merge.
Summary
Updates the deferred Bash wait test to allow slow Windows shell startup without relaxing its frame-loop check.
The Bash wait test measures responsiveness only after the wait enters.