num
How many holders are waiting for guide-path space to drain.
The cost of draining rather than evicting, measured. A closure that is wanted now and begins late because an aisle was busy is a real effect on whatever the holder represents, and it is invisible unless it is counted. Its time-average is the mean number of closures pending over the run.
Here rather than on the holders, and that is the point of this whole construct rather than a detail of it: a model may make as many holders as the run turns out to need -- an arrival stream of spills, a picker entity per rack face -- so a response per holder would be a response count that nobody can state before the run. Four numbers describe any model.