Here's the system: 64-bit Linux; BOINC 7.2.42; AMD FX-4300 (4 core); Nvidia GTX-650; 8 GB RAM; Buffer parameters 1.0 days minimum + 0.5 days additional.
--Seti CPU ID=4774476--
Abbreviations used for active projects and resource shares:
S@H Seti@home 80 CPU + Nvidia work enabled
E@H Einstein@home 20 CPU + Nvidia work enabled
A@H Asteroids@home 5 CPU work enabled

S@H configured for 3 concurrent tasks, E@H and A@H configured for 1 concurrent task. If other tasks are suspended then S@H happily fetches CPU and Nvidia work. But if A@H, for example, is running its one allowed task then S@H does NOT request CPU work and eventually the buffer CPU work is depleted leaving 2 or 3 of the cores idle.

Example situation: Feb. 20, A@H running one CPU, S@H running one Nvidia, E@H waiting to run (only Nvidia work available). Work requests to S@H were asking for ZERO seconds CPU work. O.K., "suspend A@H project" and --IMMEDIATELY-- work request to S@H and 55 CPU tasks downloaded!

A@H -resumed- and there have been no further S@H CPU work downloads for 3 days now. All the 55 downloads have been processed and reported. The buffer has no S@H CPU work and is not requesting any.

Maybe related observation -- for at least a week now, since Feb. 17, A@H has been running one CPU task (all that app_config allows) full time despite the relatively low resource share setting. It's cranking out 9 work units per day.

I've added cc_config log flags for <work_fetch_debug>, <sched_op_debug>, and <cpu_sched_debug> and what I hope will be relevant output follows:

.../snip/...
23-Feb-2015 13:50:11 [---] [cpu_sched_debug] using 1.80 out of 4 CPUs
23-Feb-2015 13:50:11 [---] [work_fetch] Request work fetch: CPUs idle
23-Feb-2015 13:50:11 [Asteroids@home] [cpu_sched_debug] ps_150218b_1138_5_1 sched state 2 next 2 task state 1
23-Feb-2015 13:50:11 [SETI@home] [cpu_sched_debug] 03se12ac.22341.4566.438086664208.12.66_0 sched state 2 next 2 task state 1
23-Feb-2015 13:50:11 [---] [cpu_sched_debug] enforce_schedule: end
23-Feb-2015 13:50:14 [---] [work_fetch] entering choose_project()
23-Feb-2015 13:50:14 [---] [work_fetch] ------- start work fetch state -------
23-Feb-2015 13:50:14 [---] [work_fetch] target work buffer: 86400.00 + 43200.00 sec
23-Feb-2015 13:50:14 [---] [work_fetch] --- project states ---
23-Feb-2015 13:50:14 [NFS@Home] [work_fetch] REC 59.786 prio -0.000000 can't req work: "no new tasks" requested via Manager
23-Feb-2015 13:50:14 [Asteroids@home] [work_fetch] REC 495.453 prio -0.355004 can req work
23-Feb-2015 13:50:14 [SETI@home] [work_fetch] REC 24465.446 prio -1.127314 can req work
23-Feb-2015 13:50:14 [Einstein@Home] [work_fetch] REC 6830.041 prio -2.041003 can req work
23-Feb-2015 13:50:14 [---] [work_fetch] --- state for CPU ---
23-Feb-2015 13:50:14 [---] [work_fetch] shortfall 6051.51 nidle 0.00 saturated 126446.69 busy 0.00
23-Feb-2015 13:50:14 [NFS@Home] [work_fetch] fetch share 0.000
23-Feb-2015 13:50:14 [Asteroids@home] [work_fetch] fetch share 0.053
23-Feb-2015 13:50:14 [SETI@home] [work_fetch] fetch share 0.842
23-Feb-2015 13:50:14 [Einstein@Home] [work_fetch] fetch share 0.105
23-Feb-2015 13:50:14 [---] [work_fetch] --- state for NVIDIA ---
23-Feb-2015 13:50:14 [---] [work_fetch] shortfall 34731.81 nidle 0.00 saturated 94639.88 busy 0.00
23-Feb-2015 13:50:14 [NFS@Home] [work_fetch] fetch share 0.000 (no apps)
23-Feb-2015 13:50:14 [Asteroids@home] [work_fetch] fetch share 0.000 (blocked by prefs) (no apps)
23-Feb-2015 13:50:14 [SETI@home] [work_fetch] fetch share 0.889
23-Feb-2015 13:50:14 [Einstein@Home] [work_fetch] fetch share 0.111
23-Feb-2015 13:50:14 [---] [work_fetch] ------- end work fetch state -------
23-Feb-2015 13:50:14 [---] [work_fetch] No project chosen for work fetch
.../snip/...

I hope to resume S@H CPU work "soon" which can easily be done by momentarily suspending A@H but I'm willing to defer that action to do any further diagnosis of boinc in its present state (i.e. S@H work request = zero seconds cpu).

Thanks for any help.
GeneAZ

Just seeing the words "concurrent task", "app_config", "work fetch", "cores idle", etc. made me think of this post:
http://boinc.berkeley.edu/dev/forum_thread.php?id=9900
May it be the same problem?

flakinho -- and anybody else...
(RH your posts to the other thread duly noted)

Yes, I did see the "...CPU threshold" thread you had started earlier. I was not sure whether my configuration (not using VMs) might give additional insight into the issue. It appeared to be boinc related and not project dependent. However, in the 7+ hours since my original post E@H has downloaded 4 CPU tasks (one of which is now running) and 3 Nvidia tasks.

Now I'm more confused! Why does E@H fetch CPU work and S@H does not? (And, yes, I have noticed that the S@H servers have been short of work during this time.) Does it make a difference that A@H and S@H projects have app_info.xml configurations and E@H does not?

There is a work-around, of course. Just suspend other tasks momentarily every two or three days to let the S@H work refill.

GeneAZ