When a fix or change breaks functionality that was previously working, causing previously-passing tests to fail.