From what I've read, it seems like you have a good chance of making this game if you play your cards right :)
The biggest issue I see is that you'd need a lot of server power for this. But the way people move around from zone to zone playing mini-games lends itself perfectly to a peer-to-peer architecture. You would still need servers to act as the main gate between areas/mini-games, like Battle.net does to help you join a Starcraft multiplayer game, but it would blend seamlessly into the whole experience of travelling around the world since it replaces the menus; and this way you wouldn't need to worry at all about servers. The only reason why servers need to control every tiny piece of physics in the game is to prevent cheating. This is only an issue in games like RPGs, where you have a kind of a virtual property and cheating can destroy that. But in a platformer, unless you have stuff like money and items, cheating will make you win a mini-game, but that's it. People might get pissed off and stuff, but the balance of the game is not lost. If you had a server-client connection that is not 100% perfect, you'd run into the same problems. This is EXACTLY what happens with games like Counter-Strike. I'd go with a peer-to-peer approach, so each mini-game would be essentialy like a lan-game, or a chat room if people are not really playing (towns and stuff) - the Starcraft approach I mentioned.
Oh, and you can still have items as prizes for special mini-games made by the designers, if these games are hosted by your servers so that it's just as hard to cheat as with a server-client architecture. This way people can't cheat to have lots of cool items with no effort ;)