Doh. I'd forgotten about the serial bootloader cause I try to never use it. The problem that HJ v394 fixed is seen as soon as the kernel accesses >32MB, which I'm guessing wouldn't ever occur during a serial bootflash.

Ok, not so serious then. I guess I could rebuild an .upgrade file containing any hypothethical software release so that it had a different kernel and thus avoid RAS pin issues. It would be a pain though as everyone with >32MB would need to do so in order to use any such release.

I'm fairly sure that you have picked up that change anyway - it's been there for a few weeks now.