That said: discussions can be about the nominal topic, or they can drift, naturally, to other areas of interest. I find the meta-topic of "where is a good place to discuss the things I'm interested in discussing" to be generally on-topic. HN isn't a bad place for that.

With the ability to collapse threads baked into the HN page design, if you're not interested in a discussion, you can simply collapse it an move on.

On which: I'm increasingly hitting long threads from the bottom, for a few reasons:

1. The top-ranked post often isn't particularly on-topic, and attracts diversions posted for visibility.

2. It's easier to tell if a thread is of little concern if you follow it from the bottom.

3. It's easier to track a conversation from bottom to top and collapse going up, than it is to backtrack back to the top and collapse that. There's a subreddit which runs the collapse bars down the side of the thread, allowing a subthread, or entire thread, to be collapsed in one go.

4. There are often downvoted or flagged items which I feel are incorrectly tagged. It's usually faster to assess if something has some merits than to see if it's solid, and I'll nudge stuff up -- even if I substantially disagree with it -- if it seems it's been overly penalised.

HN, as with other online discussions, does poorly for deep and complex posts. One standout was The Edge Question issue a few years ago. There were a handful of comments on the first few essays, but given that there were ~100 - 150 total responses, it's not the sort of thing you can digest quickly.

I've tracked down other issues of the Question and gone through it. I have to say on balance I'm fairly disappointed in the quality -- much is narrow self-promotion of research, another large set is rampant speculation. The short an humerous responses are often disarmingly refreshing. And every so often there's something quite good. Sturgeon's quality estimate though seems generous.