There are approximately 52 unadvertised /8s. Before the addresses ran out, we were using about one /8 per month. So even if we'd reclaim all unadvertised ranges, that would still buy us maybe 4 years (probably less considering the explosive growth in Asia).
The design of IPv6 began in the mid-90s, and RFCs were released in 96-98. Now, 15 years later, we are mostly still using IPv4.
In conclusion I think saying that we would have enough time to implement entirely new protocol if we'd reclaim unused addresses is way too optimistic. Even getting IPv6 really deployed at large in 4 years would be challenging enough.
Somebody explained up above, but one example is that it can be used to mitigate collisions. If you have a company merge with another, and both are using 10.0.0.0/8 you could spend a lot of time and money renumbering or you could use unadvertised IPs to connect the networks.
The design of IPv6 began in the mid-90s, and RFCs were released in 96-98. Now, 15 years later, we are mostly still using IPv4.
In conclusion I think saying that we would have enough time to implement entirely new protocol if we'd reclaim unused addresses is way too optimistic. Even getting IPv6 really deployed at large in 4 years would be challenging enough.