Building and observing virtual communities is the most fun I've ever had in my life (dabbling self-proclaimed psychologists could write dozens of blog posts on the matter :) ).
For almost 7 years now, I've been a moderator of a once very active (several thousand daily active users) and still somewhat active forum; and the mythologies, unspoken norms, cliques, memes, etc. that emerge are just fascinating. It's really a miniature world in itself.
What's even more interesting is adding arbitrary game mechanics on top of it. I've built a couple PHP webgames when I was in college, and while none of them became quite big, fascinating patterns emerged.
In most of them, I would, by design, let as many elements as possible be up to the players; and you end up with micro-societies that tend to show the same basic behavioral patterns as our own, just on a smaller scale.
For example in one of them instead of just collecting resources and spending them to build weapons like most games of the genre do, the game would force you to join a coalition where a player-elected leader would decide how to spend resources collected by the players. Players could plot to overthrow the leader, or re-elect him if they felt he was fair, but also smuggle resources to enemy coalitions, etc. In some coalitions, the leader tried to be fair and just, but that would ultimately lead to his demise; in some others, leaders would be dictators that the other players actually appreciated and supported; and in some others, the leaders would plot like crazy with some players while pretending to be honest publicly.
Yeah, these experiments are fun and humans are fascinating :)
There's not that much literature, surprisingly. You can find some basic tutorials on how to create scripts to do basic things, but to my knowledge there are no complete compendium walking through the creation of a complex game like OGame or Travian (two of the most well known) from A to Z.
As far as users, I publicized the games to my friends and on forums where I was an active member, and it was pretty much all word of mouth from there. I never went past 4 digits though.
As for what stops people playing, it's a good question with varied answers. Some people have naturally a short attention span, and move on to other things. Some people just slowly fade away after a few weeks/months because they find another newer game they like better. Some stay for a long time even if they get bored by the game because they like the community. It's very varied.
Ultimately, if you want to keep your players, you need to be active in the community you created: organizing events, communicating with players to solve their problems/questions, adding new features (and publicizing upcoming features they can get excited about); and all of that is very time consuming and hard to do if you're not full time on that project.
My first ever project was a text-based MMORPG game written in Perl. This was back in 1999, so I have no idea if there is material anywhere on making PHP or Perl games like that, but my approach was to "just make it".
That project had 10,000 active users, and it was all word-of-mouth.
