So just to wrap this up, it looks like I get bus errors in general wherever flash media are involved-- CFs in the AztecMonster, CF connected with adapter to an R-IDSC-E/R, and SD connected with bridge board to an R-IDSC-E/R. What doesn't give me bus errors are MicroDrives attached to either of those two. Nuking the contents of the flash disk and reinitialising it does seem to keep the bus errors at bay for a boot or two, but they always return. It's extremely frustrating since evidently a great number of people run their SE/30s from CF with no such trouble, while I can only use storage that spins.

I had a potentially-related error with a Soekris (open-platform router) box loaded with OpenBSD. If you did a hard poweroff after a crash when booting from CF, the system would consistently and repeatably fail to boot afterward until it was PXE booted into a skeleton OS and the CF's volumes fscked manually. With a MicroDrive in the CF slot (and an extra IDE-SATA bridge plugged into the HDD IDE interface to slow down bus polling enough to give the MicroDrive time to init and be recognised by the BIOS) the system would consistently and repeatably boot and execute the expected auto-fsck, then complete its init and come online.

I conclude that CF doesn't handle errors elegantly and isn't suitable for my application. I'm sick of this issue, and I'm just going to use MicroDrives, since they work. Feel free to correct/educate/flame me.