There is a corner case that can happen especially if you are using partitionable slots.
If there are no other users in the pool who have jobs that match the requirements of the slot, then
the low priority user will keep on matching to that slot. We see this effect because most of our priority
users run 8-core jobs while the opportunistic low priority (high prio factor) users run 1 core job.
As long as there is a 1-core slot out there and the user who requested 1 core job has jobs in the queue
then it will keep on matching to that slot, even if CLAIM_WORKLIFE is short. Only solution is an aggressive defrag.
Steve Timm