In response to: http://lesswrong.com/lw/c1/wellkept_gardens_die_by_pacifism/

I'm a moderator at Elitist Jerks (http://www.elitistjerks.com), a World of Warcraft discussion forum. Within the WoW community, EJ has always been known for its strict moderation standards. We're exactly the sort of 'well-kept garden' that EY's post is about. You can see the fruit of the mod team's labor here: http://elitistjerks.com/f34/ I'll give some of the site's backstory for non-WoW players, describe the crossroads that we're currently at, and then give some caveats before you generalize too much from our example.

EJ's initial community came together to discuss WoW's most challenging content, known as "raids". In order to optimally outfit our characters for maximum performance in raids, both empirical and theoretical work was necessary: the game's combat mechanics were reverse engineered and detailed models for each character class were created. Within a couple of years, this "theorycrafting" work became the forum's primary purpose - refining and updating models as new game patches were released. Throughout the forum's life, high moderation standards have been maintained in order to protect our high signal/noise discussion. Primarily, asking for help is forbidden when the resources to answer your question already exist.

However, we're starting to wonder if we've performed our task too well. 

• Discussion about any aspect of the game that can't be quantified has almost completely died out. Despite having threads explicitly designated for more subjective topics, the overall forum atmosphere is sufficiently hostile/intimidating to non-analytic personalities that those threads don't see much traffic.
• Further refinements in class modeling are so small as to be empirically unverifiable - their impact emerges in lengthy runs of simulated combat, but is washed out by random factors (critical hits, etc) in the 5-10 minutes of a typical raid encounter.
• The community is heavily dependant on a few dozen key contributors who maintain the accepted spreadsheet/simulation models. By encouraging all our visitors to use their tools, we've made these people into single points of failure. When one of them quits the game, it's difficult to find another to take up maintenance on their model, because any programmer would rather write new code than work on old code (http://www.joelonsoftware.com/articles/fog0000000069.html).
• In addition to the previous, some of the class models simply aren't as good/reliable as others. However, due to the average person's inability to shut up and multiply (http://wiki.lesswrong.com/wiki/Shut_up_and_multiply), we don't allow posts questioning model output unless they have a detailed proof of a bug. As a result, the models that should be taken with a grain of salt are treated with just as much credibility as the strongest.

So here we moderators sit on our porch, having kept our garden tidy for six years now. The questions we're asking are "Is this the community we meant to create?" and "What happens to a community formed to solve a problem once the problem is effectively solved?"

Caveats:

• There is a private area of the forums where subjective and off-topic discussions still take place. However, these discussions are invisible to non-paying members. I personally worry that quarantining our fun may be just as dangerous for the health of the community as diluting it. This opinion is not universal among the moderation team.
• WoW is, in the larger scope of things, not that hard a problem. What to do after accomplishing a primary goal is not likely to be a problem for a more broadly-scoped community. Even in WoW, hard problems remain that resist quantification - How do you motivate 25 people to keep battling a dragon that's been killing them for the last 2 hours? How do you identify the recruits that will best fit into an existing group and its culture? How do you balance redundancy and responsibility in leadership?

edit: fixed some link formatting