Flaky Test

A flaky test is one that yields different results — sometimes passing, sometimes failing — without any change to the code it exercises. Flakiness erodes trust in the suite: once a failure might be “just flaky”, real regressions get ignored.

Common causes include reliance on wall-clock time, ordering assumptions, shared state between tests, network or filesystem access, and unsynchronized concurrency. The fix is almost always to remove the hidden source of non-determinism rather than to retry the test until it passes.