I've had this thought too. It'll be very difficult to pull off, though.
If you go for the "one leader/coordinator, everyone else contributes", then it's practically no different to any other fangame being made. The "contributions" will only be vague suggestions, meaning no actual work will get done except by the leader, who will inevitably give up once they realise they're stuck doing everything by themselves. Also, the community will never agree on anything (plot/pacing/Pokémon/mapping/etc.), so no progress will be made there anyway unless someone (the leader) takes charge and defines a direction - in which case, again, it's no different to any other fangame.
There is an alternative, though, and it's one I've been thinking about and which may have potential. One person works on the game at a time, and possession rotates between those who volunteer in turn. There is a strict time limit of 1 week per person, during which time they're allowed to do anything they want to the game (including undoing/redoing other people's work without remorse). At the end of their week, they send a download link to the next person on the list, and post a summary of what they've done for everyone to see. If they fail to do either of these things, their efforts are ignored and the previous week's version gets passed on instead (both the revised game and the summary are required, as it's a team effort and there shouldn't be secrets or hidden information between members; otherwise their work cannot be accepted).
The advantage of this is that people can have a go at a bit of game-making for a week, and will only contribute if they're actually interested enough to volunteer (which means they will do something). The time per person (1 week) is kept relatively short to encourage actual progress, rather than letting it get bogged down with designing new-for-the-sake-of-new systems (not to mention keeping things going relatively smoothly if someone vanishes). Meanwhile, the thread for the project will have discussion about the latest week's changes, collaboration about the story and so on, and will help direct the suggested progress of the game - the current week's developer still has absolute free reign over everything, of course. In the end all the revisions and possessions and discussions should come together to form an actual game.
Obviously someone needs to start things off first, perhaps before even making a thread for this project (so there's something to start from). Or maybe you could say that the Essentials example maps are the starting point, rather than starting with a blank slate like normal, and for starters the aim is to expand/turn them into an actual game (plans can always change later). I don't know whether it's better to let everyone download every week's revision, or whether it should be passed privately between each person as their turn comes round (the project's coordinater should also keep every revision for archiving, which is their only purpose). Everyone should definitely be kept informed of all the changes made each week, though.
Could this work?