Read-write cap, alongside the sibling swL2QOSSchedulingMaxPkts (both range 0-255 in the same scheduling entry), on how long this hardware priority queue can hold packets before the scheduler is forced to service the next lower-priority queue — a weighted round-robin fairness bound so a busy high-priority queue can't starve lower ones indefinitely. Note: the source DESCRIPTION for this object literally reads 'Indicates the mechanism of QOS scheduling,' which is copy-pasted boilerplate and does not describe a latency value; this reading is inferred from the object's own name and its parallel role with MaxPkts.