I think the major difficulty with user-contributed content on the web is that we don't have many good ways (yet) to deal with the volume of participation that happens on the web. Things start out fine with a few people, but once the user base gets beyond a thousand or two, it becomes much harder to maintain the sense of community or engage in valuable discussions.
Please excuse the extended references to the Other site, but I think there are lessons to be learned from what's happened to tOs.
I've been on tOs since the early days (I'm user 1451), and it's been interesting to watch the way the site changed as it got popular. In the early days, it was not uncommon to have discussions on a topic that extended over quite a few days and brought out interesting points later on in the discussion. The signal-to-noise ratio was reasonably high, perhaps because people were there because they wanted to be part of the discussion.
As the site became popular, the number of less-informed posts increased, and that made it harder to find the useful posts and engage in discussions. Putting the moderation in place helped for a while, because it made it easier to find some of the valuable posts. Granted, sometimes the moderators make mistakes or introduce bias, but it tended to work anyway. The meta-moderation seems to have dealt with mis-moderation reasonably well.
As tOs grew, the default changed from seeing the full text of every post to only seeing posts rated 2 or above, after the number of posts reaches a certain threshold. Once an article reaches this threshold, you only see the title of the post -- and the titles are rarely indicative of the content. Replies get treated even more harshly: you never get to see the content of a reply unless it's rated 2 or above. While this did help reduce the volume of posts that people had to browse through (and probably reduced the load on the servers), it also made it harder to find the occasional good post that starts at 0 or 1. This also added a burden to the moderators: in order to be effective, they now have to actively examine 0 or 1 posts based only on the title in order to find the underrated good posts. From the moderation that I'm seeing right now, few moderators make the effort to find a couple good posts among the hundreds of posts that start at 0 or 1.
Once the article reaches 100 posts (in an hour or two, usually), anything posted at 0 or 1 is rarely read by anyone else, unless a conscientious moderator seeks it out and moderates it up. This makes it hard to hold extended discussions, and also limits the participation to those who have constant web access. I used to be able to hit tOs once a night and get in on good discussions. Now, it's mostly a read-only forum for me rather than a community that I participate in.
Why spend all this time discussing tOs? Because the lesson is there: unorganized volume is the enemy. Once the volume of user-contributed content reaches a certain point, it's very difficult to find the valuable contributions. Without some kind of method to separate the useful information from the rubble, it's hard to have good discussions.
Interestingly enough, this is the same problem with the web -- too many pages, and hard to find the useful information. We're starting to see useful solutions in this arena, though: Google's search system really helps me find what I'm looking for on the web. Of course, weblogs are a different problem domain, so they need a different solution.
If you haven't figured it out yet, I'm very biased towards anything that promotes discussion. Here's a couple of ideas that might help out as kuro5hin gets more popular -- it'd be sad to see it fall into the Slashdot Morass (TM).
Order the articles by amount of recent activity, and perhaps the overall rating of all the posts under that article. This will keep the more interesting discussions near the top of the list, and people will be more likely to join the discussion.
If posts are collapsed to show only their title, then randomly expand some of the posts to allow people to read and rate them without having to explicitly choose to follow the link. This will help to bring out the posts that have good content but don't indicate that in the title.
Share a post's ratings with its parent by giving the parent a bonus based on the children's ratings. If a reply that's down a level or two gets a higher rating, then it usually means that all of the parents of that post might be worth looking at, also. This will make it so that sorting the top-level posts by their rating will bring interesting threads to the top as well.
Share a post's ratings with its children by giving the children a bonus based on the parent's rating. If the parent is worth reading, then it's likely that the children might be worth reading as well. This bonus shouldn't be as much as the previous one, though, because it's more likely that a good reply means a good initial post than the reverse. This will help to bring out responses, perhaps introduce differing opinions, and hopefully provide more fodder for starting interesting discussions.
I think that the best way to attract the good participants is to set up a system that rewards intelligent discussion. We can do this by making it easier to find and participate in. The trick is finding methodologies that will do this, and being ready to implement them as the critical volume is reached.