I don't see any thermal compound residue from the VRM heatsink on the MOSFETs (SweClockers link). Seems to me you *really* need some sort of thermal interface, given that there may be some microscopic variations in height from MOSFET to MOSFET. Wouldn't surprise me if that blown one was a couple thousandths of an inch lower than its neighbors (and not touching the heatsink as a result).
Edit: Well, OK... at least there's a compliant thermal pad on the underside of the heatsink. Still seems non-optimal to me, given the length of that heatsink. Even slight warpage of the PCB or heatsink is going to result in less contact pressure on the MOSFETs near the middle of the VRM area. Note that the blown one is indeed towards the middle, and the pad seems to be indented less towards the middle of the heatsink as well.
One of the links also mentions possible issues with the VRM controller firmware. This seems plausible to me; you'd kind of have to work at it to blow a MOSFET like that on a regulator design with *18* MOSFETs to share the load!