fix(test_wait_tool_early_end): scope the wait-tool fake clock to the wait loop thread

[tlobinger]

Summary

This PR combines 1 related change:

• Scope the wait-tool fake clock to the wait loop thread

Changes and rationale

Scope the wait-tool fake clock to the wait loop thread

Problem: test/test_wait_tool_early_end.py fabricates a fake monotonic clock by patching the GLOBAL time module (patch.object(_time, "monotonic", ...) + patch.object(_time, "sleep", ...) in _run_wait, ~lines 62-67). The patch is process-wide, so any background thread alive in the pytest worker that calls time.sleep() advances the fake clock too. The tests then assert EXACT equality (clock.t == float(MIN_WAIT) i.e. == 60.0), so a single leaked 1ms sleep from an unrelated thread fails the test.
Why it matters: Every full-suite run on a busy host can fail on this module, burning local gate attempts and CI reruns for unrelated PRs. It already burned one ship attempt of a pr-maintenance loop (zero signal against that loop's change).
Tests: Regression pinning the leak: while a wait runs under _run_wait, a helper thread calls time.sleep(0.001) repeatedly; clock.t still ends at exactly float(MIN_WAIT). Fails on the current code, passes after. Full existing test_wait_tool_early_end.py suite green: .venv/bin/python -m pytest test/test_wait_tool_early_end.py -o addopts="" -p no:cacheprovider
Review focus: backend

Review map

Change	Primary files
Scope the wait-tool fake clock to the wait loop thread	test/test_wait_tool_early_end.py

Commits

• ebd232b fix(test_wait_tool_early_end): scope the wait-tool fake clock to the …

Validation

Local build, static-analysis, and test gates completed before publication. Upstream CI and review checks on the current PR revision remain the authoritative merge signal.

Changed files (1 files, +51/-9)

• test/test_wait_tool_early_end.py (+51/-9)

This description updates automatically as related changes land; manually written text outside this block is preserved.

[github-actions]

GPT 5.6 Review — ✅ no blocking findings

GPT 5.6 completed its review of ebd232bade6f177e7b5592bb189a7b579bfab7e9 and found no blocking issues.

This comment is updated in place on each push.

Review details

No findings.
[GPT-REVIEWED] ebd232b

False positive or not applicable? A repository writer can comment:
/ai-review override gpt ebd232bade6f177e7b5592bb189a7b579bfab7e9: <one-sentence reason>

[github-actions]

Opus 4.8 Review — ✅ no blocking findings

Reviewed ebd232bade6f177e7b5592bb189a7b579bfab7e9 — this comment is updated in place on each push.

Review details

This is a test-only change. The _Clock captures the owner thread's ident and the real time functions at construction (before patching), then routes non-owner threads to the real functions. The wait loop runs synchronously in the owner thread (confirmed by the existing exact-equality asserts that require the fake clock to advance), so the scoping is correct. The new regression test starts its background sleeper before _run_wait, so it exercises the non-owner path as intended.

No correctness, security, or resource concerns on the changed lines.

No findings.

[OPUS-REVIEWED] ebd232b

Verdict parsed from the review's SHA-scoped output markers for commit ebd232bade6f177e7b5592bb189a7b579bfab7e9.

False positive or not applicable? A repository writer can comment:
/ai-review override fable ebd232bade6f177e7b5592bb189a7b579bfab7e9: <one-sentence reason>

[github-actions]

Design Review (Fable 5) — ✅ PASS

Advisory design-level review of ebd232bade6f177e7b5592bb189a7b579bfab7e9 — updated in place on each push; does not block merge.

Design-Verdict: PASS

Root-cause fix at the right layer: the fake clock now engages only on the wait-loop's own thread, with a regression test that pins the leak.

[DESIGN-REVIEWED] ebd232b