However, I'll hypothesize that it's preferable (overall) to have more friction on collapsing comments. A lot of great discussion happens in response to even a mediocre top-level comment, and forcing the user to take a few extra moments to skim the child-comments will increase the serendipitous discovery of these nested gems. Furthermore, the problem of a mediocre top-level comment with mediocre nested comments being at the actual top of the discussion is mitigated by the ability for people to downvote the thread, as has been done (with varying levels of efficacy) until today.
Excellent phrasing. As a fairly long-time user who lurks more than he comments, I can concur that much of the best conversation I've read and participated in happened to be such nested gems. Serendipitous, indeed.
Is this actually data-driven or just a guess on your part? My experience is that a mediocre comment rarely ever gets good replies.
Maybe "mediocre" is the wrong word, as it implies an absolute judgment (on some non-existent scale of "quality"). I guess my intuition is derivative of the "don't judge a book by its cover", in that even if a submitted and upvoted story is not something I feel like reading, I still might read the comments to see the discussion.
Maybe someone internal to HN can do a quick calculation of how many child-comments surpass their parents (while accounting for the differences in visibility).
This design was the minimum change possible (which has a value, but usually a net negative one).