flaky

A flaky test is one that sometimes passes and sometimes fails on unchanged code, making its result untrustworthy.

Flakiness usually comes from hidden nondeterminism — timing races, ordering assumptions, shared state, or randomness — so the same test gives different verdicts on different runs. Flaky tests are corrosive because they train people to ignore failures, masking real bugs. Fixing them means removing the source of nondeterminism, not just rerunning until they pass.