Back in the End Times of Google+, I had access to a data dump of all n-million Google+ Communities metadata (name, subscribers, and visible activity on the Community homepage, the last of which gave an indication of post cadence).
I noticed two odd things:
1. Taking a log-log plot of number v. size of Communities, there were far fewer very large communities than a straight power-law relation would suggest ... by a factor of 2--4 or so. This didn't look extreme plotted log-log, but when you looked at the raw values, there was something dampening the count pretty markedly.
2. In terms of activity, posting cadence fell off once a community hit (very roughly) about 10k members. Above that there still might be a lot of activity, but it was a lot less than size alone would suggest.
I've never been able to conclusively answer why either of these behavious should be the case, though I suspect that:
I suspect that the first has to do with the tendency of very large forums to go rogue (be taken over by bots / shills / spammers, or otherwise go off), to be moderated by site admins (excessive and pervasive violation of site policies), and/or simply put real people off participation because there's no genuine reward for doing so (questions aren't answered, or are responded to with incivility and/or spam).
The second (more activity in mid-sized communities) would probably relate to the last factor above: with ~10k members (give or take an order of magnitude), there's probably a core of a few hundred to a few thousand active participants, which is about at the threshold where associations and affinities can form. People still have conversations, at least part of the time.
Keep in mind that G+ (similar to FB) has a flat discussion format, so that the most recent comments are appended to the end of a post. Reddit, like HN, has a threaded discussion. For a long time I criticised G+ for not supporting threads, but I came to feel that this was actually a strength where discussions were well-moderated, as there was a sense of a real discussion occurring, where later respondants would often refer to an earlier comment (usually naming the author and quoting a bit of context) in replying. Conversations could and did last for days, weeks, months, and years. By contrast, on Reddit (and HN), a conversation is effectively dead within a day or so, if not a few hours. Comments are simply howling into the void.
This happens somewhat less on HN than Reddit, largely due to the finite size of HN's front page (infamously: 30 stories, though many might cycle through over a day). Odds are good that a top-5 story actually has some exchanges on it, and I've carried on a few of these for a few days.
But I'd long ago noted that Reddit was where conversation, in the sense of an exchange between multiple parties, goes to die, especially on any of the larger (million-plus follower) subreddits. Smaller (~50 -- a few thousand members) generally does far better.
In a similar sense, HN's quiet days (weekends, holidays, or during The Burn) often see far less activity but far more interaction.