It happens when metrics suggest that a system is well tested, but important behaviours, risks, or failure scenarios remain ...