I am still using my T100TA myself, but it's too giving me a few problems (made no better by the fact I run Linux on it), including some of the ones you mentioned, but not this particular one.
However... just an idea... I don't remember what it was exactly, I have trouble focusing, but there was *some kind* of Intel "security" (or rather, user-disabling) technology whereby a coprocessor would perform a sanity check on something that was running on the main CPU, and if it didn't pass, it would shut the computer down. This wouldn't happen immediately, but after a preset timeout. I know I'm being very vague, I don't even remember which Intel chips this applied to, but shutting down after an exact amount of time does sound like a self-check triggering. I will try to be more specific if I can recall the thing I'm thinking of, with the help of some googling.