kevinveenbirkenbachandClaude Opus 5.5 40681927de test(test-e2e-playwright): prove run_parallel overlap without racing stub jobs ...

The concurrency tests of run_parallel read overlap from wall-clock stamps that include each stub job's python start-up, so under the 80 xdist workers of make test the jobs serialised by accident (quality-high on store-i18n 1f15f528b3: test_onion_jobs_share_tor_with_each_other 1 != 2; on 1c922d39f3: test_a_freed_slot_is_refilled_without_waiting_a_poll 0.6508 not < 0.6). A stub job can now wait until another job has started, so the overlap the tests claim holds by construction and a serialising scheduler times the waiting job out. The freed-slot test polls every 30 s and bounds the refill at 15 s: the green path never waits a poll, and a scheduler that does misses the bound by twice.

🎯 Fix: 94 %
💥 Break: 1 %
💸 Cost: –
⏱️ Time: –
🛡️ Security: 100 %
🏆 Score: 93
🚦 Gate: ✅ 90/5, cost and time <= 100, security 100 %

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

2026-11-08 13:39:23 +00:00

..

__init__.py

…

test_run_parallel.py

test(test-e2e-playwright): prove run_parallel overlap without racing stub jobs

2026-11-08 13:39:23 +00:00