I’ve been testing with it powered from the bench PSU today (albeit without the -5V, but IIRC that’s only required for audio/serial/cassette).

As far as progress, I removed IC42 (destructively – cutting pins) and socketed a spare video ULA – now I’m getting the output clock signals that I’d expect to see, and the CPU is receiving a 2MHz clock. Initially RST was stuck low on pin 40 of the CPU, yet was going high everywhere else (tube socket, cartridge slots, etc). Closer inspection showed some noise on the RST signal at 2MHz, which made me suspect some sort of connection between pins 39 & 40 on the CPU. A clean of the PCB in that area resulted in RST going high on pin 40 as expected – so perhaps that “dust” on the board is more conductive than it looked… or something else was going on that cleaning somehow resolved.

A quick probe around the CPU showed it was trying to do things – address lines were active, and it was hopping in and out of ROM. MOS ROM was being selected periodically, but it sort of looked like the CPU was running amok all over RAM for the most part. This led me in the direction of a potential RAM issue (thinking being that if RAM was pooched and zero page wasn’t reliable, then the CPU could be off doing literally anything)… I was getting RAS at 4MHz just fine, but CAS was stuck high on IC26 & IC18, but seemed to be OK on IC23 & IC17.

CAS for IC26/IC18 is generated separately from IC23/IC17 (not sure why yet, but it is). Essentially pin 11 of IC34 is stuck high, so tomorrow I’ll get the logic analyser out and see what pins 12 & 13 of IC34 are doing (there’s signals on them, but I’m not sure if they’re coinciding in the high state, which would be required to make pin 11 go low). Perhaps this particular gate in IC34 is a dud (but two others are used in generating CAS for the RAM chips that are getting it, so if this IC has an issue it’s not affecting all the gates).

Edit: reading the 4464 data sheet, CAS staying high /might/ be normal if the machine is using RAS only refresh and the memory concerned is otherwise not being accessed… More research required.