Alright, well given that we're using the same PROMs from the PCB now, and the 3rd match palette doesn't corrupt on the PCB, then yes, it seems a MAME bug, maybe the theory about the palette bank being used is correct. Thank you for testing, it's good to have the original reference.
You are welcome!I can confirm the BPROMs dumps are good, I dumped the devices three times with different programmers and I got always the same CRCs
The extra rows of cheerleaders look intentional based on my own study of the code, there's code that loops around based on the number of goals you've scored, up to a limit of what can be shown on screen. I can't give caius an easy test for this, because the code does seem to have additional checks (involving the protection device) against modification, even if you match the checksum or patch out the initial checksum.
I put all my effort into it but, me bad, I was not able to score a goal:(
Anyway, I can keep the board still for some time so feel free to ask about further checking