I'm not sure how well this would work in practice. The reasons for the previous changes were that performance of the existing line had reached a dead end, and the new line was noticeably better right from the get-go. Not the case for an Intel->ARM switch.

Though - whether this would be a problem in practice, I couldn't say. Most software probably doesn't demand much. GPU-heavy stuff wouldn't be much affected. Emulated Photoshop would probably be terrible, but major vendors would get advance warning, I'm sure, so you'd have ARM-friendly Photoshop as a launch title.

And, better still, AArch64 is little-endian, and I think it handles misaligned data transparently (and if not, Apple could presumably fix that?) - so for 90+% of software, "porting" to this hypothetical future ARM OS X would require little more than a rebuild.

This wouldn't help Boot Camp much, though... I use Boot Camp quite a lot. Windows compatibility played a large part in my decision to buy a Macbook Pro in the first place! But I wonder how common that is?

Rosetta worked, but I don't know if I'd say it worked well. It was brutally slow to the point of being unusable with some apps on mid/low end Macs.

Their solution for Mac LC was effectively an Apple IIe shrunk down to a PDS expansion card.