It would be possible to do basically this with the existing IPv6 header and much of the v6 infra. All existing IPv4 hosts get the same /32 on v6, and any NATs stay intact. On the UI side, keep the ipv4 address format like 8.8.8.8. Going to V6 is very low-risk, low-impact this way. V4 packets can even be converted to V6 and vice versa because it's the same address space at first. Once there's sufficient adoption of V6, individual network operators can break things up into /40, /48, etc if needed, which wouldn't be reachable via V4 anymore*. Back to UI, a /40 would look like 8.8.8.8.4.
Unlike today's v6 which is a totally separate network with new routes and randomized addresses. It also has v6-mapped-v4 addresses, but that's different.
* If you really want some more temporary compatibility, /40 would still be sorta reachable from V4 if you added a special kind of NAT that encodes the extra 8 address bits into the port number, but idk if that'd be palatable.
Unlike today's v6 which is a totally separate network with new routes and randomized addresses. It also has v6-mapped-v4 addresses, but that's different.
* If you really want some more temporary compatibility, /40 would still be sorta reachable from V4 if you added a special kind of NAT that encodes the extra 8 address bits into the port number, but idk if that'd be palatable.