These are mostly happening on a machine that is shared by World Community Grid and scheduled by BAM. I'm thinking maybe that's the cause of the problem and that I should switch things around so I'm not running multiple projects on one machine. Then again, it might just be that I'm caching too much work. What do you think?

"Average turnaround time 10.02 days" = you seem to use ~10 days WU cache
http://asteroidsathome.net/boinc/host_app_versions.php?hostid=208165

And the project have ~10 days deadline (judging by the diff "Sent" -> "Not started by deadline - canceled")
http://asteroidsathome.net/boinc/results.php?hostid=208165&offset=0&show_names=0&state=6&appid=

BOINC should work on these in-danger tasks in "High Priority mode"
(Not shown by latest BOINC Managers, shown by BoincTasks in Red)

Why BOINC didn't move those tasks on the front of queue I don't know.
Maybe you set 0+10 days WU cache?
Change to e.g. 1+5 days (the first number is not only WU cache, it also tells BOINC to finish and report tasks 1 day before the deadline)

---

- ALF - "Find out what you don't do well ..... then don't do it!" :)

Okay, thanks! I took Asteroids@home off that A8 rig and took World Community Grid off my i7 laptop, so now nothing is shared. It used to annoy me that the scheduler would stop an A@H model that was almost complete to start a WCG from scratch. I don't think I ever saw an Asteroids task running in high priority mode although Einstein@home ones frequently did that! Have taken your advice and set the laptop to 1+5. Now I just have to watch that the project isn't out of work units for too long.

Maybe you set 0+10 days WU cache?

It was actually set at 2+1