This is more true of communities built around ideas rather than activities, and there are also moderating tactics you can use to avoid it. For example, online communities based around astronomy, or radio-controlled airplanes, or baking, or other hobbies remain relatively healthy even decades after their founding.

Also there are communities around conventionally caustic topics (eg. Lambda: The Ultimate for programming language design, Penny Arcade for games) that manage to survive for a decade+ with no loss of quality because of moderation policies. For example, L:tU has an "avoiding ungrounded discussion" policy - every thread must be centered around discussion of a published academic paper, which first of all keeps the focus on people who actually do work, and second of all discourages everyone without the background to read and understand academic papers.

This is what has killed Twitter. Both social justice Twitter and right-wing Twitter revolve around competing for status-points, and backslapping (the latest trend is linking to other people's egotistical tweetstorms along with a comment like 'This' or 'Thread'.

Very little challenge of ideas takes place - either from within communities or from outside. The whole thing has become a Wittgensteinian language game where people compete to praise or denounce the latest political trend in the most caustic or verbose way. It's tiresome and offputting to people unwilling to subjugate their entire online persona to their identity.

I think you have outlined the basic issue with human social interactions of any size greater than 3 people.

> the people who have the most time and energy to invest in discussions (which eventually snowball into flame wars) are the people who aren't actually doing the actual work.

Pretty sure we've reached peak-discussion. Future online communities will be based primarily on activity.

Interesting analysis, seems to explain my personal observations.

Only wrinkle I'd propose: The "moderates", or any of the characters you define, aren't fixed, and personal circumstance often colors which role one plays at any given time.

A very interesting point of view, what do you think can be done to counter this outcome.

Github managed to escape that dynamic up until the last couple of years.