Field note /
A regression test that only correlated with the bug
Upstream review changed what the regression evidence could support.
CorrectionExternal upstream · Codex-LBPR #2506 · under review
How the claim changed
Belief
Repeated failure on old code and success on the fix appeared to reproduce the shutdown defect.
Evidence
The proposed upstream fix was accompanied by repeated measurements.
Contradiction
Upstream review showed that the trigger was correlated with the race rather than causally synchronized to it; an unbounded post-cancel wait also remained.
Correction
Bound the post-cancel wait and replace the timing-sensitive trigger with a deterministic barrier.
Current claim boundary
A regression test is stronger when its trigger forces the claimed failure condition. The upstream pull request remains under review.