distributed protocols can also be popular. E.g. torrent or email
Not sure they're good examples tbh.
Most of the torrent ecosystem still relies on a few big trackers and indexes; few people actually go DHT-only. And the popularity of private trackers is even more so an example. Also, as soon as DDL/Streaming became popular, they eclipsed the torrent scene, as far as normies as concerned.
Email's benefit is that everyone keeps an archive of things. But as a tool, protocol, and UX-wise, it's unthinkably shit, really. We just happen to not have a much better alternative right now. But in practice people truly only use it as last-resort, because it always works ™️, not because it's any good.
And even then, a handful of companies (Google's Gmail/GSuite and Microsoft's O365) account for the vast majority of email traffic, with few people actually running a mail server (because it's ridiculously hard to do correctly).
The truth is that nothing fully decentralized will even gain traction, because of the operational effort which isn't fun nor worth it for most people. The best you can get is probably a ActivePub-like level of decentralisation, where people will end up signing up on a couple of large instances which peer with one another.
Well, until something like Threads comes up, and 90% of the public uses it as their home server. Whether it still is decentralized at that point is up to semantics. You'll argue that as long as the protocol allows it, it is. Until Threads (or whatever other biggest player) writes their own extensions or supports only part of the protocol. Just like Gmail did with IMAP in the past. Then it's decentralized-originally-but-defined-by-one-corp.