Two clocks run during Linux md RAID recovery: the array must regain redundancy, while production requests still have latency objectives. The safe target is not the highest rebuild number. It is the fastest rate that keeps workload latency inside an agreed budget without leaving the array degraded ...