Well, in the whole process I did discover one interesting thing, even though it did nothing to solve the problem:
When using an opt-24, the cardcage pcb is actually not needed. It seems to be enough to simply connect the opt-24 to the clock card.
This is nice, since I removed the whole cardcage except for the pcb.
Anyways, the opt-24 has three adat i/o ports. (3 in 3out).
Only one pair isn't working - the other two seems to be fine.
The rest of the HDR system seems to be fine - the meters show activity both on the front and in the software.
Only, no output or input on that one pair.
However, clock signal IS sent over that pair.
Again, if I remove the opt-24, and instead connect one opt-8, in the "offending" position, I get random clicks and glithces - very typical for clock problems - but the glitches are only there on those specific 8 channels - the others work just fine.
To me that points quite a bit at the clock card.
If I understand the service manual correctly, one ribbon cable is inputs, the other outputs. If it was a bad ribbon cable, I believe it would just be either inputs or outputs that didnt' work. But here the problem is all eight channels on that particular adat port - both in and out.
So by process of elimination, I believe the problem is around the clock card.
The problem is the same, whether I connect to the d8b or to my motu 1248, regardless of which device is set as master or slave. (In my experience, if the problem is clock glitches, then the problem will be on all channels, but I would still get normal audio with all the glitches. Here I get NO audio out or in).
I've cleaned the contacts on the clock card before, and don't think they would get bad over a few months. But to be certain, I'll try cleaning them.
By the way, I've also tried moving the clock card (and also the opt-24) to other pci slots, and everything is just the same.
Good thing is it seems the clock card and opt-24 doesn't necessarily have to be in the first two slots, so this could give a bit of cooling space for the graphics card. But again, this indicates that the problem is not with the motherboard slots - again pointing to the clock card having a malfunction.
(Maybe I'll try smacking the machine with a chicken filet while I'm at it, just to try something.....)