@Uplinger: Please consider upping the limit for max no of tasks outstanding per core, currently set at 25. I think that limit was set at a time when the length of FAHV WUs was very unpredictable, the average length was longer, and occasionally this was causing WUs to miss their return deadline.
Now there will be no more for a while of the faah WUs to pad out the cache length, and FAHV WUs are much shorter and much better-behaved, so the limit makes a joke of the work cache setting in our BOINC clients.
At the start of yesterday's outage, my machines were maxed out on WUs per core but some machines started running out of work after only about 14 hours.