I was pondering various things last night while I lay awake in bed, and I remembered MUQ, which I hadn't thought about in a while (a language/server/persistent programming environment for developing MUDs). It seems like it got a long way, developed by one person who really wanted it to succede, and then for whatever reason his interest faded off. Maybe his ambitions were accelerating slightly faster than his progress, which is discouraging.
Even though it's aging as we speak, it still seems like a more thoughtful and robust solution than I've seen before. A not-so-ambitious project built on it (like a shared world with tile-based graphics or something) could be quite effective. It would still require a lot of work from the author, as well as others... another aborted effort probably wouldn't be useful, though.