Are you measuring temps on the GPUs after the reboot or before? Temps can fluctuate wildly within 10 seconds of stopping a CPU/GPU stress test. You might try running the game in a window while watching GPU temps via your tool and CPU temps in CoreTemp. See if one of them goes over the manufacturer's specs or close to it. If that doesn't yield anything, you might try running a memory test app like BurnTest, Prime98, etc. to see if you are rebooting from a memory corruption error instead of a temp issue.

Most CPUs these days will self-throttle their performance rather than rebooting spontaneously. I think the same thing applies to modern GPUs but I haven't verified that.