So, the unit I have that I'm having issues with seems to just lock up. cgminer is still running as the elapsed time, local works and discarded keep going up.
I'm not near the unit to verify if it is actually hot or if the temp sensors are reporting incorrectly, but if they are reporting correctly it would appear that the unit is still at least dissipating a normal amount of heat. It's just not submitting any shares.
Restarting cgminer fixes the problem.
Known bug currently, sorry about that. Soft (usually) or hard reboots will clear it. In terms of frequency, I had this happen once in the first 24 hours, and then never again.
When they do lock up like that, are they chips still doing work? The temps reported were the same as when the unit is under full load, so it would seem that it is unless the temp values stop reporting at that point as well.
The old Gen1 Avalon minis would do that as well. I could deal with a miner crashing and going down, but those stupid things would crash and sit there burning 600W until they were power cycled. They're the reason all my gear is on networked PDUs now.
Note: After the cgminer restart the unit ran for 40 minutes and then crashed again.