Hmm. Strange.

I knew there were 4 or 5 devices on that bus, I just wasn't sure which used which lines. From all of the datasheets, it looks like the USB and Crystal chip only use D0-D7. The SMC chip and IDE header both utilize all 16 bits. Not sure if anything else is on the bus.

Unfortunately, I did completely remove the SMC chip and the unit is still acting up.

I'm fairly certain that it's not an issue with the IDE cable as this drive and cable work perfectly in my MKII. I suppose it could still be a poor connection of the header to the board, but it would have to be a broken pad or trace or similar. I verified every pin with my multimeter. My thought is a damaged trace from the IDE header back to the StrongArm? Can you tell which lines are acting up now? I wish I knew more about CPU engineering and busses to troubleshoot this stuff myself, but I don't quite yet. I've found a few websites like this which have been helpful, though. If you can tell which lines look like they're acting up, I can try to do the same thing as the Crystal chip and run a jumper from the StrongArm right to the header. Perhaps I should verify continuity from the IDE header back to the SA to verify all 16 Bits as a start?

EDIT: No continuity back to the SA from the IDE header. Seems they're probably buffered.

EDIT 2: I found a handful of 74LVX4245 transceivers, no doubt matching the 3V logic of the SA to the 5V HDD. Alright, then...no direct bypassing.