Application of thermal paste and mechanical assembly including mounting boards to heatsinks was not Imet's responsibility. All soldering meets IPC class 2 and inspection was done by trained personal. Everything was in order when delivered to AMT. The likely cause of your boards not working is overheat. The heatsinks used on the top of the ASICS are not adequate to dissipate the required 30 percent.
Yes the top sinks were woefully under sized. However, bet the rig you posted as running fine has the same tiny sinks, no?Just that the final flaming nails in the coffin are the missing thermal bumps on back side of the board that are supposed to complete the thermal contact path to the main heat sink. I refer you to my previous post on this https://bitcointalk.org/index.php?topic=569769.msg7610781#msg7610781
I have sample boards ranging from no thermal bumps at all to most all there. On average I see what is shown in the above post - most missing meaning that there is no way in hell to make decent contact with the main heat sink as the few bumps present act as jacks raising the board up from the sink. Thermal compound no matter how carefully applied (or drunkenly slathered on) will not compensate for the 8mil gap. So much for the 70% of heat flow path.
AMT was not responsible for those bumps. The CM was. If they were not explicitly called out in the BOM from AMT/Bitmine (my copy makes no mention of them nor having filled thermal vias) then either have none or all of them. Period. If you have ever worked with power devices before you *have* to had known that. Given that some/many bumps are there tells me you knew about the thermal loads the A1 has. After that
Then there is the dpot chip substitution without regard to the core voltage divider reference circuit it controls... Ohm's Law applied to parallel resistances anyone? Frankly, aside from the legal who-did or didn't do-what points, this is all just stomping around in the rotting carcase of a long beaten to death horse.