Been thinking this through. Need cake, and to eat it too.
Discussion
If this were to work it would be a godsend for most people.
Goal is for minimal relay impact, zero protocol impact, monster client added capabilities, and migration toward dedicated key management software/hardware rather than leaning on clients for that at all.
NIP-26 would have been fantastic but essentially no clients wanted to support it. So minimal impact with minimal changes are they easy to go, IMO, if possible.