I would want to go even further, and strike out (perceived) "importance" as a barrier. Thinking in terms of "importance" will tend to cause our minds to stay within certain topic clusters, when what we actually want is more variety of topics. Rationality lessons are often most illuminating when applied in situations we don't stereotypically think of as illustrating rationality lessons. People may have pet topics or specialized areas of expertise that they would like to post on, but don't because of a fear that their subject isn't "important enough" (which in practice tends to mean being about the topics most commonly discussed here). This is unfortunate, because rationality literally applies everywhere; and I think an aspiring rationalist should seek out as many diverse opportunities for honing their general rationality skills as possible. This will prove useful when it comes to the "important" topics.

On the other hand,

These have already been discussed so they would be discouraged as duplicates rule (except for substantially new approaches),

I actually wouldn't want to restrict duplicates to new approaches to the subject itself; I think a new specific lesson on rationality should suffice. Familiarity has its advantages too. (For example, there are a number of Bayesian lessons that I have learned from my study of the Knox case since the original discussion, and I would hope to be able to post in the future on some subset of these, using this particular vivid illustration, without too much objection on the grounds that the topic "has already been done".)