Once during testing, I had a pod that dropped a chip and cgminer started to misbehave, which pegged the CPU use. I mentioned the bug to kano and his preferred fix was "stop using hardware that fails". That's about the only time I've seen one of these guys not running smoothly so could be related to your issue. My test guy ran 40 days uptime with no problems before a power outage forced an unplanned restart.