I suggest everyone to think for a moment about the fact that Eliezer somehow created this site, wrote a lot of content people consider useful, and made some decisions about the voting system, which together resulted in a website we like. So perhaps this is some Bayesian evidence that he knows what he is doing.
There's also plenty of Bayesian evidence he's not that great at moderation. SL4 was enough of an eventual failure to prompt the creation of OB; OB prompted the creation of LW; he failed to predict that opening up posting would lead to floods of posts like it did for LW; he signally failed to understand that his reaction to Roko's basilisk was pretty much the worst possible reaction he could engage in, such that even now it's still coming up in print publications about LWers; and this recent karma stuff isn't looking much better.
I am reminded strongly of Jimbo Wales. He too helped create a successful community but seemed to do so accidentally as he later supported initiatives that directly undermined what made that community function.
Seems to me there are two important factors to distinguish:
To me it seems that the problem is the inherent difficulty of herding cats; and Eliezer is the most successful example I have ever seen. I have seen initially good web communities ruined after a year or two... and then I read an article describing how exactly that happened. From outside view, LW seems to survive for surprisingly long time...
If it's worth saying, but not worth its own post, even in Discussion, it goes here.