So why not just shift over to IPv6? Because the same arguments and economic analysis my advisor and I made in 2000 are still true. IPv6 does not replace NAT; IPv6 instead requires NAT for compatibility with the rest of the world. Given the choice between IPv4 + NAT and IPv6 + NAT + IPv4 there is not a whole lot to recommend the extra layer. (Especially since IPv6 fails to solve some of the real scalability concerns of the Internet...)
It's true that non-U.S. deployment of IPv6 may continue to grow, and software support increase. Heck, my laptop configures its interfaces with IPv6. (It wouldn't be the first time technical communities charged off in the wrong direction.) But I think address transfer markets will help clarify the real cost of IPv6 transition, by putting a fair price on globally routable addresses. If there is a lot of market demand for globally routable addresses, it will drive IPv6 adoption. If addresses are cheap, nobody will bother.