I wonder if LLMs could be useful here for automatic node-graph generation of which replies addresses which train of discussion within a thread, and the user can click through said generate index to follow how a specific topic evolved.
Many chronological forum software also can already display reply/replied-to chains (though perhaps not first-class in terms of UX) if people use the reply function, which is often an option.
I built this (both chronological view and new comment filtering) into the comments presentation on https://hcker.news. Check it out, I’d be interested to know if there’s any way I can make it more useful.