if we look at social games, they are iterating every single day. It's come to the point where Zynga can tell if a game will be a hit or not, purely based on the collective data that they've achieved, over the course of 4 years. Not one Zynga game has bombed since Mafia Wars.
Some people will be turned off by the earliest form of the game, but there are always more clicks to be had and leans startups using MVPs have to deal with this too. The key is to be learning about how to improve the fun factor every release.
As the game continues to improve, the user base will continue to grow. I'm not totally convinced that this could only apply to mmorpg's, but I'm willing to concede that it seems like the best genre to attempt this and to apply the microtransactions model.
The problem is that most single-player games are finite. Once people have completed them, they're done. Replayability is something comparatively few people can stand for any length of time, especially in a crowded market. Add the problem of an exceptionally short tail for most games (20-50% of total sales in the first month would not be atypical) and the value of updating the mechanics of your single-player game becomes much less valuable. Best to make another game in the series that improves all those flaws.
A few exceptions, all indie games: Minecraft, Eufloria and Darwinia. All of these have had major updates to gameplay over time.
That being said, I wonder if games might be prepped for disruption. A big part of any polished game is user testing. How do we know its fun? We watch people play and see if they have fun.
Well, with the advances made in web analytics and the possibilities that graphics acceleration offers in modern (or maybe future) browsers, I could see studios and publishers selling games around a more iterative process.
An approach that finds the minimum fun experience and iterates from there could work very well on the web. Games like Asteroids and Missile Command do not need the polish others describe to be engaging, and then hosts of new features can be added over time to make such games more engaging. Add in the fact that web-based business models for games have proven recurring revenue through micro transactions can be very viable, and I think there could be a lean approach to game development.