Interpret suite reliability when retries and quarantine change the report.

Fictional case: in week A, 500 fixed test cases run once each on each of 20 builds, yielding 10,000 case-build observations. There are 9,600 first-attempt passes and 400 first-attempt failures. Each initial failure receives exactly one retry on the same build without code changes: 320 retry attempts pass and 80 fail. In week B, the team quarantines 40 of the 500 cases, removing them from the merge gate. The 460 remaining cases run on 20 builds, yielding 9,200 case-build observations: 9,016 pass first attempt and 184 fail; one retry recovers 160, while 24 still fail. The quarantined cases produce no week-B results. Product changes, environment load and case difficulty are not controlled across weeks. Compute the rates, then assess the claim that the full suite became more reliable. “Passed after retry” is an observed outcome, not proof of a harmless test-only cause.