Just btw: I know that you checked the soldering, but connector stuff is always a little tricky.
Check the voltages on the 24AA. The supply and the ground lines. Perhaps something is switched.
Also, the 24AAs seem to make lot more trouble than the 24LCs. I also used AAs in my SID and I switched them all to LCs, because random sounds were coming out (!).
Although this issue was solved by the hardware change, so really *do* consider doing it.
There must still something be wrong with your hardware. As PayC wrote, he noticed similar obscurities with his 24AA256… maybe you should try different cable lengths or so… thats all what I can say to this topic.
Have you also tried a “MIOS BankStick 32k” Request and Send with SysExBox?
to give you some more technical backgrounds: in mios_bankstick.inc of the MIOS source code you will find the MIOS_BANKSTICK_Write_Poll loop with the note “memo: time out counter should be inserted here!”
This is the only point where a BankStick transaction can hang, and the reason for this is either a damaged EEPROM, or an unstable connection. So, definitely a hardware failure (which is not handled by MIOS yet, but for which I don’t see a strong requirement)
Another point of interest: mb64e_bank.inc, search for “MB64E_BANK_WriteBankStick” you will find following comment “reset core if bankstick not available anymore (just for debugging)”
This case can happen if the BankStick is disconnected during a write operation, or when you’ve a bad junction.
So - both cases explain what you’ve noticed, therefore no bug, but a deterministic behaviour (from the software point of view)