Find hidden coupling among tests

Inputs: [TEST SUITE], [SHARED FIXTURES], [FAILURE PATTERN]. Map mutable shared state, cleanup, clock and randomness. Propose shuffled-order and isolated reruns before changing production code. Output: A coupling hypothesis list and minimal test-only changes. Check: The suspected test fails or passes predictably under isolation. If essential information is missing, identify it rather than filling the gap with a guess. Run the suspected tests alone and after neighbours to expose ordering effects.