The reset logic can be complex in that any number of system devices can cause a reset. The outputs of U141 are open collector, so anything else sharing the _EHLT _CPURST or _RESET lines is capable of causing a system reset, including the CPU board.
I'd expect U140 to have a totem pole output, so therefore the lines _FPURST _IORST and _ERST are only driven from U140 outputs. That's why these lines don't have pull up resistors, while the lines on the U141 outputs do.
Not sure exactly what drives the U150 internal _KBRST logic, no way of easily knowing what goes on inside the gate array. I'd imagine _FAIL has some influence on the _KBRST line. Aside from that, U150 is mostly doing address decoding.
If you've measured _FAIL at 5V, it's probably OK.
Quite possible there's something dodgy happening on the CPU board, as it's capable of generating system resets. The best suggestion is to try a known working A3640 or whatever to isolate the fault to the CPU board or main board. Else you could be chasing your tail for days.