I decided that rather than start reflowing or checking all the connections between the major chips, I would just capture data off every pin on the Glue, MMU, and Shifter. I can do that pretty rapidly and then compare data between like pins. As anticipated, the Glue was producing 15 kHz sync, despite the system being booted in mono. Didn't take me too long there to notice that RW on the Glue was stuck high, where it was quite active on MMU. This led to the discovery that RW on Glue was connected to the new 2.2k pullup, but not to RW on MMU, Shifter, or 68K. After looking at the board to figure out what had happened, I eventually realised that there was (past tense here) a tiny little trace between pin 9 on the 68K and the pullup resistor right next to it. Looks like I managed to break that when removing the old pullup.
So, a quick replacement for said trace on the underside...
Boot it up...
And back in business with a correct mono display.
