Andrea Cifftolilli's Phantom Authority case study on Wikipedia applies team and good club theory and transaction cost analysis to provide insight into how large scale wikis work. Do read this paper...I'm only providing a single point here.
First the numbers, one of my minor obsessions: Wikipedia has currently an unknown number of anonymous contributors and 29,853 registered users of which 143 are administrators and seven out have developer rights. This plus one founder that plays the role of "benevolent dictator."
Unlike Slashdot, the prototypical example of an open self-organizing social software community with low administrative overhead, Wikipedia doesn't use an explicit reputation system. Instead, it functions at two levels. The first level, that of proceedural authority, gives users the ability to contribute and edit (away graphitti) at low transaction costs. The second level, institutional authority, is given to administrators.
About 150 Administrators...there's that number again -- a ceiling of cognizance for a social network. The eight people, a core group with developer's rights, is what some developers consider to be an optimal team size and within the boundaries of a creative network.
So behind the scenes of a successfully scaled community the empowers users (for horizontal information assembly...more on that later) is an active social network that relies on social practices that are not hard-coded or codified. The paper suggests that to scale further a reputation system may be required for this network, a major change to manage given the culture that reflects and drives its tools. Perhaps it should look for a set of new challenges to hand off to a new group of administrators to delay such a rash transition.
