I found myself thinking a few days ago that half of threading on Twitter could be implemented with simple microblogging and a protocol for back-reference URLs (all of which can be done in something as simple as hugo or Jekyll) coupled with a standard for the user-agent to "unroll" the thread. This requires neither a centralized or federated service; it's built on basic web tech.
The only piece of the puzzle missing is that it only works one way; to see the whole 'conversation,' you'd need to find the tail of it (most recent comments). This sort of data-mashing is one of the things a centralized server can do simply and well (though it's certainly implementable on a federated service). But without a federated service, a rule for semantically identifying "topics" and web-crawler that knows how to sort by newest-to-oldest on a given topic identifier gives you the whole story.
[0] https://indieweb.org/reply [1] https://indieweb.org/Webmention
I hope once a Fediverse culture takes root we will see a second generation of federated services that start experimenting and doing things in ways that are, for lack of a better term, "natively" federated rather than simply trying to mimic something closed. That might be when it really takes off, when you start seeing things that work best in a federated environment and make awkward fits for a centralized one. A good example of something like this might be the mod scene in gaming, which Steam tries to get a piece of but can't quite seem to corner.