Possibly noise on the power or ground lines? Clock line noise? Are there any capacitors on the mpu that are across the power line to filter noise that may have gone bad ? Pull-up resistors on bus lines or tied off control pins that may be bad or missing ?
No schematic in front of me right now to consider other things...
To debug noisy signals we used to put very high value caps (for high frequency noise ) soldered one lead to ground and then touch the other lead to different signals like a clock or the bus lines.... see if that gets through test cycles more consistently .
Some random thoughts...