Yes, you got it right. Weird, right? My only guess so far, from 15 years of experience overclocking, watercooling, messing with HSF's and such is that when you run a device at 70-80C for long periods of time the TIM either changes its properties or gets so thin that it "squeezes" out slightly and when the chip cools and contracts it lets some sort of micro gaps form and you get hot spots which crashes the "chip". That is just a guess, though. Now that I think about it that doesn't make much sense because the only reason it should crash is if you see the die temp get too hot... Maybe I have worn the transistors out from all the heat/current and they only run right when there is full power flowing through them (or more specifically-full voltage, seeing as modern chips vary the voltage automatically with clock speed)? Here is an up to date article that explains what I learned a few years ago about how transistors wear out:
http://spectrum.ieee.org/semiconductors/processors/transistor-aging/0I'm on my 3rd power supply. This one is an 80plus Gold OCZ 850W and I only pull about 550W.