The chip itself has an overheat threshold which is around 130's C. If the chip shuts down, the chain breaks cause the whole the unit to basically shut down from a hashing perspective. If you're using putty, you'll see an error message similar to CMD_READ_REG ERROR. If you see this, unplug it, wait 30 minutes restart it via putty and do the following:
root
amt
killall cgminer
cgminer -DT
And check to see if the cores in each chip are ok.
Also there are two 4-5 fans in this case. There are two 140's under the cage blow air up through the fins. The 200mm exhaust on top and the 120mm x 1 or 2 in the front of the case blowing air in.
So...if you have 2-4 restarts like the highlighted one and you are looking at 1-2 hours of downtime daily plus that you need to monitor the thing the whole fucking day. AMT you must be very proud of your work.