The criterion is simple: a small number of reposts is ok when an article is solid and hasn't had significant attention on HN. We want good stories to have multiple cracks at the bat, or sheer randomness would prevent many from ever getting seen. If you pay close enough attention to HN that those two previous submissions counted as already having seen the story, you're way beyond the 99th percentile!

That's why we leave the dupe detector so porous. HN's goal isn't to build a database of unique urls—it's to have the most interesting front page possible.

This system is easy to abuse, so we ask people not to abuse it. For example, don't delete things and then repost them. But I don't see abusive reposting as a huge problem on HN right now. We catch most of it, and alert users catch nearly all the rest.

Edit: I should mention that we're going to be releasing a new approach to /newest that addresses the tendency of good stories to lose their chance at the front page too quickly. Exactly when we release it depends on how much code I get to write in the next month or two.