Thanks for sharing your repair experience.
It looks to me that there are 2 legs in that circuit, either of which could cause this problem.
One is a real voltage drop in the path involving the photocoupler, FET2 and the bleed resistor.
The other is an error (thermally-induced) in the voltage measurement leg feeding into the ISA215 hybrid board used to sense the HV.
Since the issue seems to be thermally induced and is not a permanent failure, and you thought replacing the op amp on the hybrid board fixed it (in the measurement leg), then either the series resistor string or the little chip capacitors would be good suspects. i hope he replaced the caps along with the resistors.
Those little ceramic chip capacitors can get stressed and crack at the solder joint, then get leaky and recover in thermal cycling. The little chip resistors could change value under thermal cycling, but it is a less common failure in my experience.
They put 3 caps in parallel, likely as a redundancy against failure, but there should have also been a series string to be effective. Tesla does this on their battery cell voltage measuring circuit.
kiev = kenny's innovative electric vehicle