Have seen an interesting failure mode on the SMPS power supplies at my old workplace.
During testing, we had one power supply burn up after being subject to excessive load for 48 hours (below the current limit trip at high-line voltage of 253V AC). We couldn't figure out what happened but essentially the phenolic PCB and the MOSFET had got very very hot and burned up a huge section of the board. The MOSFET was shorted and the fuse was popped, but as far as we could tell, that wasn't the cause of the burning.
Later, we were able to replicate the failure. Under high temperature the power supply controller could overheat and fail, but when it failed its output didn't go to zero. It sat around 2-3V, which was provided by the Vcc charge circuit (the MOSFET was not switching at this point.) So the MOSFET was being continuously biased with low gate voltage when it was supposed to be off, so it was conducting current continuously. This wasn't enough current to blow the fuse, but was enough to make stuff very very hot and start burning stuff in the area. Eventually, the MOSFET would succumb to short-circuit failure due to overheating, and pop the fuse, but that wouldn't happen for some time. I wonder if something similar happened in your case.