I think the switchover point from A to AB would be determined by the initial bias setting, and not the components per se.
(Note I know enough to be dangerous)
I think you'd want to trace the runs on the board and determine what the trim pots are adjusting. Running it way out of whack for any length of time will kill it proper.
And I agree that components should be replaced in pairs, or a total replacement across the board (if possible). Some amps, like my Counterpoints, are not tolerant of replacements, as they had to be tightly matched from hundreds to get tens, long discontinued MOSFETS...