Newer drives, especially SSDs, are a lot faster with ATA protocol state transitions than older drives. This exposes flaws in the original (Intel) ATA protocol spec, as well as races in the Linux IDE code that I wrote nearly 20 years ago.

I believe those races have been discovered and fixed in upstream kernels, and I could probably find/fix them here with a few hours of fussing about.

There's also the possibility that the protocol bridges themselves may require slightly special handling, though I doubt it. Those things are not well tested on ancient, slow, PIO-only hosts (like the empeg) though.

Cheers