tomich, what is your cache size set at? If you are running a lot of HCC WUs and your cache size is set close to 7 days (the deadline for HCC), then it will force WUs to high priority. and if a lot of WUs go high priority, boinc may switch between them.

.....confused! Set it much LOWER, say 1.5 days, that will sort it!

That ''puzzling'' behaviour is being revised, debated, revised, debated and revised more with the currently developed 6.12 client. Presently, BOINC 6.10 appears to compute a HP result for a while, determines it's going to be okay, then finds that another result slipped into deadline thread, so switches again and again and if lots of projects have this and LAIM is on or off, more memory is eaten and it ever getting worse, and all because of high caching and substantially variable run times. Those that do non-deterministic calculations just cant be accurately TTC projected. Just the last day, the quad with 2.5 day cache inflated to almost 8 days because 1 task took 3 times longer than estimated. BOINC assumes all following tasks will also run 3 times longer. Would that device have had HCC in the queue the s$%t would have hit the fan, but they were not, so in about 20 further tasks the cache normalized again. In that the solution offered is simple as offered by Scribe: Set that cache to 1-1.5 days and there's virtually no chance of this situation developing in the first place.

Happy crunching.

edit: 3x, not 4

LAIM= Leave Application in Memeory (When pre-empted)
TTC = (remaining estimated) Time to Completion

AS for #2, downtime is very rare, updates that are going to cause outages are usually posted in the "Known Issues" thread and normally only last for an hour or 2

system going down due to failure have all been fixed in a few hours, I think the worst was about 6 hrs

I keep my cache at 1.5 days and have never even come close to running out of work