I think Anatoly's proposals have merit.
We certainly don't want to allow the changes to impact our culture in a negative way. Cautiousness is warranted, because wikipedia works well as it is. We're just looking to the future and anticipating how to scale as we get more popular.
I'd just like to point out that I dislike "point" systems because they explicity avoid simple human subjective judgment. They're just bad AI. The problem we're trying to solve is inherently one of subjective value: most people's edits are valuable to us, but some aren't. I don't want to abdicate responsibility for that judgment to a stupid automaton.
If we want to give "reliable" users privileges that we don't immediately give to everyone, that's great. But just give them-- don't build a whole system of automating it, just give them to those who ask, and who have earned them. I would far rather that flag be set by a real human being exercising judgment than by some meaningless process.