Considering that second asic to be defect and all other ones working, would mean that the excess power (5A *3.3V) has to be dissipated in that second ASIC. I don't think that 16.5W additional power dissipated in that second asic, will only result in a few degrees temperature difference in comparison to these other three ASICs. I think, that all ASICs are pretty much dead, and the second ASIC only has more internally circuitry blown and shorted out, thus the lower resistance.
As there are no other hotspots on the board and the only hotspots are those 4 ASICs tells me the following:
As this low resistance can be measured in both directions it's likely that there is a real large short in those ASICs, where maybe a huge part of the Die is shorted out.
As the current decreases while the ASIC heating up tells me that there is no semiconductor material involved, so there could only be a complete short between the some bond wires and the metal Base- plate of the die. Because if a semiconductor would be involved the current consumption will increase, because of the negative temperature- coefficient of semiconductors. As the current decreases with increasing temperature this short must be "metallic" because only this would explain the positive temperature- coefficent.
Although there is a second possible alternative: The 3v3 power plane, the ASIC pins and internal bond wires are heating up, increasing their resistance and and thus lowing the current consumption, although there is a internal semiconductor short.
A possibility is it's not the asics at all, but something they are driving, pulling too much current through logic outputs.
That would be possible, but this also means that those logic outputs maybe defect forever, because of overcurrent.
Dave can try desoldering this 2nd asic, but that would be tricky, because of the large heatsink, which can't be removed, because, as he said, it's very well glued on that ASIC. The only possibility to remove a such high pin count device, is with hot air. Because of the huge thermal capacity of that ASIC with heatsink, that could be a hopeless undertaking, except Dave will be able to remove the heat sink.
Cheers Gregor