Decide borderline readings by protocol, not by mood
Critical because: 31% of rejections were unnecessary — over-cautious receivers guessing on readings like 8.5 °C surface instead of probing or escalating.
SME: named failure #2Frequency: every borderline loadObservable: yes (rejection note field)
Practiced in moduleBorderline incidents A–C: core 8.4 °C → reject protocol; missing logger + driver claims → hold-and-escalate; each choice plays its consequence before feedback.
Certification7-of-8 randomized scenario bank, pass ≥85% (6/7 = 85.7%) — every item is a borderline call, none is recall trivia.
Job aidDecision table: in-range / out-of-range / no-data × accept / reject / hold-escalate, laminated at the goods-in desk.
Built-in mistake (distractor source): rejecting a good load "to be safe" — module shows the ripple: dock backlog, driver detention, cost write-off with zero risk reduction.