Maybe try checking the CMOS battery. Does it POST reliably?
I have the older ancestor of this exact embedded Atom Supermicro
board, the A1SRi-2758F. It's the second one I've had; the first fell
victim to the infamous clock bug on the Atom Avoton/Rangeley series.
This one is an RMA'ed one that is supposedly fixed.
It worked fine from when I got it back from RMA until a few years ago,
and then it would do exactly as you describe- just randomly lock up.
Eventually this progressed into refusing to POST reliably.
Re-flashing the firmware over IPMI would fix it but it would always
regress after a time, especially if power was removed. Until one day,
I realized the CMOS battery was reading low in the sensor values.
Replacing it resolved all these issues- though the voltage still reads
low in IPMI.
Since these two boards are such close ancestors, I wonder if the
successor board still shares this same quirk, assuming it is one and
mine isn't just broken. They certainly look almost identical from
photos.
