The problem is fundamentally content overload, but that goes hand-in-hand with decrease in average quality. Quality, in an absolute sense, may increase, but it's overwhelmed by ignorance and mediocrity. So there needs to be filtering on the content lifecycle.
Filtering right up front is very effective, but bad for business. Take Google Answers. You had to pay money (mostly) to ask a question and few were allowed to answer them. Extremely high average quality content for what was there, but where are they now? Compare that, of course, with its polar opposite Yahoo Answers. It was never even August there!
But the internet's native model is to eliminate any constraint on generating content. That's why Eternal September was so grievous -- it's when that philosophy got mugged by the reality of induhviduals, spam, crazy people and the sheer volume of mundanity.
It was also when people realized that the potential of the internet was rapidly approaching. But okay, given Eternal September as a reality of content generation (don't try to set up walled gardens), what do you do? You bust out the best tools you can for dealing with it -- the PageRank, the groups, the tags, the voting, the social connections. And people seem to enjoy the internet again. And I think some money was made along the way.
So is it more interesting that Eternal September was the end of the golden age for the elite or the birth of a golden age for the mainstream?
MetaFilter is a classic example of reasonable but comparatively high barriers to registration: $5 for an account and you have to wait a week to post. Seems very clever, but I'm not a MeFi regular, so I'm not sure how well that's worked out.
I say that paid barriers don't do that much good or bad themselves based on the fact that on MeFi they seem to have been an asset whereas on, say, Something Awful they seem to've just turned the place into an echo chamber, and on Kuro5hin all it's done is drive the site further into stagnation. Paid barriers aren't risk-free either; it makes account gaming even more fun and makes your site a juicier target for attacks, especially if it means you end up storing people's personal details.
What's surprising is that some of the group dynamics that contribute to the break down of online communities was identified in the oddest of settings decades ago...
"On February 9, 2005, AOL discontinued newsgroup access through its service (this was announced on January 25, 2005)."