Cross-run flaky detection now marks only the non-reproducing failure as flaky. A passing run that proves the same test case can succeed on the same commit and scheme is no longer counted as flaky. This fixes inflated flaky_run_count, flakiness_rate, and auto-quarantine thresholds that could trigger from a single retried flaky commit. Flaky-group views still display the full pass/fail breakdown for the commit by fetching the complete run history while identifying the flaky group from the failing runs.
Hive
Flaky run counts now count only failing runs, not every run in a flaky group
Published
Jun 23, 2026 · 16:27 UTC
Repository
tuist/tuist