I have a feeling that the tree structure of comments in which most of the discussion usually takes place under the first few most upvoted comments might not be the best to have scientific discourse, but also don't know what would be better.
On the other hand, papers/journals themselves could be seen as a ultra-high latency social network in which replies happen in the form of papers that reference the work they're replying to.
The high latency could be seen as a feature. The ability to post a reply instantly and without much thought definitely degrades quality.
Magically? Jokes apart, it'd be something like Pythagoras posts a link to his theory, with threads going until a paradigm change prompts another link and thread, and so on, and then Newton, and Max Planck... you know, implementation details.
On the other hand, papers/journals themselves could be seen as a ultra-high latency social network in which replies happen in the form of papers that reference the work they're replying to.
The high latency could be seen as a feature. The ability to post a reply instantly and without much thought definitely degrades quality.