And they're back! It seems that all that is needed to get these two trend graphs working again is 24 hours. In both cases where CPU load and users/processes trend graphs have stopped working they have started working again 24 hours after stopping. The 24-hour periods began at different times of the day * one at about 3 pm local, the other at about 5 am local.
There were no reboots or restarts of server or client required in order to get things working again. I can't correlate any tasks scheduled to run on those machines that might have caused the graphs to start working again.
Does this offer any clues that will help us resolve this problem?