MMORPG - the role of the tutorial and self-help in complex systems

I recently started playing a new computer game - Eve Online - a massively multi-player online roll-playing game, or MMORPG for short. The game is a spaced-based mixture of adventure, commerce, pirate-hunts, and character development, set in a galaxy far, far away. The game is rich, complex, and involves interacting with real players around the world to achieve common goals.

The game is FANTASTIC! I love this style of game. But that's not why I'm writing about it...

The complexity of the environment and the rules of engagement make it almost impossible to simply document in a user manual. The item database itself - the things you can buy, find, build, install etc - runs into the hundreds of pages, and a lot of the contents won't be relevant until months after you start playing.

The problem for the game designers, and new players, is that there's so much to know and learn and yet you can't force people to spend a couple of weeks poring over a user manual before they can start playing; you have to provide an 'in' to the early levels of the game.

The game designers (and I don't think this is unique to this particular game) have tackled the problem of how to introduce new players to such complexity in two interesting ways:
i) A fairly extensive tutorial that leads new players step-by-step into the environment. From how to configure a ship, to trading & commerce, to combat and moving through space.
ii) A rich online chat built into the game that provides general how-to support for new players, and the opportunity to communicate with fellow players in real time.

With the growing prominence of rich internet applications - now in three flavours - and the increasing richness (ha) that derives from these interaction environments - I'm starting to see the need for a similar approach (i.e. an introductory tutorial) to Web applications. Whilst user research will uncover the primary tasks and objectives of the audience; and usability testing will uncover barriers to use; sometimes it will be necessary to provide a step-by-step run-through of the complex processes before users will 'get it'.

Unlike computer games, however, Web applications lack the initial commitment from the user that would make such a personal investment likely prior to actual use. So is there a limit to the complexity we can introduce into the interaction design of our Web applications before the up-front investment in time will be prohibitive to use?