Maybe just don't? Let it be IPv4 with more bits, the software is already there so dual stacking isn't so bad, adoption might actually be quick if people didn't have to learn much to implement it.
It'd be compatible (or trivially close) to most of the software that is required to make it work. Which is to say the cost of implementation would be low - not the case with ipv6.
Then it's not IPv4 and is not compatible with IPv4.