Matt : I think proposal #1 and #3 are a lot better than #2, and #1 is my favorite.
I see two problems with proposal #2.
The first problem with proposal #2 is that, as we see in democracies, there is often a mismatch between the people conscious vote and these same people behavior.
Relying on an intentional vote made consciously by miners by choosing a configuration value can lead to twisted results if their actual behavior doesn't correlate with their vote (eg, they all vote for a small block size because it is the default configuration of their software, and then they fill it completely all the time and everything crashes).
The second problem with proposal #2 is that if Gavin and Mike are right, there is simply no time to gather a meaningful amount of votes over the coinbases, after the fork but before the Bitcoin scalability crash.
I like proposal #1 because the "vote" is made using already available data. Also there is no possible mismatch between behavior and vote. As a miner you vote by choosing to create a big (or small) block, and your actions reflect your vote. It is simple and straightforward.