I've run parts at 300C for many months and seen them drift
not at all, until the oven went bad and ran away cooking the
experiment altogether (as in fire extinguishers and angry lab
techs). You need much more to drive dopant distribution.
Maybe less, to simply activate "stragglers" in the implant
population, but this would be minor.
A device without a good contact barrier system could see
aluminum or gold in-migration, if the device baseplate temp
is significantly exceeded locally. This, more than average
temp, is a concern. Averages mean Jack. It's all about the
"micro-climate" when you're talking first failure.