test(verify): gate on regressions against the generic slot, not fallback rate #179
2026-08-31
test(verify): gate on regressions against the generic slot, not fallback rate
A source-fallback percentage stopped being a readiness signal once bincraft learned to keep a matching-minor generic binary out of a fallback's shadow. What survives now is the case where the generic binary was built under a different minor: unsafe for this client anyway, so serving source is correct, just slow. Failing on that share would block slots that are in fact ready - amd64/noble sits at 53% for 4.5 and 4.6 while regressing nobody. Gate on the thing that decides it instead: packages this client would receive as source through per-minor routing while the generic slot holds a binary built under its own minor. That is strictly worse than not routing, and must be zero. Fallback share is still printed, as context rather than a verdict. Measured zero across every reindexed slot and minor.