Apologies, I've been flat out all weekend.
So, I've been looking at the pwr good signal. It goes from the in on the AT power connector, straight to pin 80 on the 82C201, marked Reset 1. This stays low for a second, then goes high.
It also goes to pin 1 on U115, an HCF4069UBE. The corresponding output of that goes to pin 13 on the HD146818, which is chip enable.
Reset on the KBC is tied to reset pin 49 on the 82A203.
On the chipset data sheet, it says pin 11 of the 82C201 should receive a reset signal from the KBC, forcing a reset signal to the CPU.
If no KBC is installed, this is floating, the board runs. If the KBC is installed, this pin is pulled low permanently, holding the board in reset.
If the KBC controller is installed, but pin 3 is lifted (I only found this out as I accidentally bent it 🤣), then pin 11 on the 201 is driven high and the board runs. The 27,25 post code comes up straight away, but it takes noticeably longer for the beep code to be output.
I need to figure out exactly how this is all being driven.
Andrew

One failed 74LS07 not supplying a clock signal to pin 3 on the KBC later and we have major progress!
Still very flakey, but I'm pretty certain thats just down to these awful single wipe DIP sockets.
Andrew

Another update. All the bigger (0.6") DIP sockets and 4x of the smaller ones replaced with round pin sockets tonight. Still not working right, but less flakey now. It always fails in the same way.
But now that I'm pulling the sockets off, I'm realising just how bad they are. From the top they look pretty clean, but underneath the legs are, in a few cases, just crumbling.
Still waiting on the rest of the sockets, but every one will be replaced.
Andrew

I had some issues with FDC not working properly with certain CF cards. This was mostly (only?) on UMC based Super I/O, even though it has 2 chips, none is a dedicated FDC or HDC, both seem to have some ISA logic and other stuff in them. I prefer Prime 2C cards, way less issues with those and usually the serial port drivers are separete DIP chips so it's possible to replace those if something fails.

If you need my help with anything, please PM me. I'm rather busy now and not visiting here as often.

Deunan wrote on 2022-03-06, 23:30:

I had some issues with FDC not working properly with certain CF cards. This was mostly (only?) on UMC based Super I/O, even though it has 2 chips, none is a dedicated FDC or HDC, both seem to have some ISA logic and other stuff in them. I prefer Prime 2C cards, way less issues with those and usually the serial port drivers are separete DIP chips so it's possible to replace those if something fails.

If you need my help with anything, please PM me. I'm rather busy now and not visiting here as often.

I have tried 4 or 5 different multi IO cards as well as a known working IBM FDC card borrowed from my 5150, complete with working cable, drive and disk. No luck at all, it either shows a read failure, or complains about not being a bootable disk.
I've been incredibly grateful for your help so far, and I don't want to pressure you, I know the feeling of being busy. I'll do some more diagnosis myself and keep this thread updated.
Andrew

Floppy drive not working could be due to I/O or DMA channel 2 not working properly. I'd expect I/O issues to cause BIOS to throw FDC failiure error rather than read error though. And this could also be down to mis-configured BIOS settings but reading the first sector for bootloader should work on any kind of drive in the same way (sector 1, head 0, track 0 is pretty universal).
So, if the drive does spin up briefly I'd look into DMA. Swap the chips around to see if it's not a fault in one of them. If not then it could be a problem with either DMARQ / DMACK signaling between ISA slots and mobo, or the chipset sockets causing trouble again - the 82A201 chip handles most of DMA state, but the bus tri-stating and actual data transfers are handled by other 82A20x chips.

So, Another update on this. I have finally finished replacing sockets on this board and its made a lot of difference. Its no longer flaky at all.
However, I think the last issue I was stuck with was some potential BIOS fault. I would get a "shutdown error" on power up, a diskette read error when trying to boot from a floppy (I've tried many different combos of drives and controllers) and when trying to boot from an XTCF card it would start to boot, before locking up.
I've swapped in a Quadtel BIOS designed for a IBM 5170 and suddenly it works perfectly, other than not seeming to like 1.44MB floppy drives. But its still progress!
Andrew