Testing should be stopped when the predefined exit criteria are met, which typically include coverage metrics, defect density, and schedule constraints. Stopping based solely on project manager decisions, budget depletion, or finding no more defects is subjective and does not guarantee quality.