KS2012: Supporting old/oddball architectures, tool chains, and devices

Indeed. For a while there was a bug in the 'alternatives' code that would cause 386 and 486 processors to crash because it patched an instruction that was already fetched into the pipeline. (Later processors will flush the pipeline in this case. These differences in behaviour are documented, by the way.) I very much doubt that, say, QEMU simulates processor behaviour down to this level - that would tend to make it much slower.