>It is not a viable option for secure communication
Security from whom? This seems like a textbook threat model problem, and one that I'm not sure how to address as a proponent of federated/decentralized services.
I used to (and still kinda do, but times change) run IRC/TS3/Mumble servers for MMO guilds. At least in this specific MMO, DDoSing was a significant issue since gear was lost on death - an attacker finds a rich target, gets their IP somehow, fights them, knocks them offline, takes their stuff.
Most/many IP leaks were via out-of-game communication services. Since a guild's Teamspeak is probably administered by a guild member, it is trivial to pull connection info on a given user. At this point in time, Skype was still largely p2p and it was fairly trivial to obtain the IP of a call participant, so it was also a big vector. There was an IRC network "for" the game, so the situation there was better than Skype+individual guild audio chats, but opers getting bribed to leak IPs was unfortunately common.
All of this meant that when Discord showed up on the scene, a lot of people in this community went "finally, a communication service where I can both 1) talk to people I don't necessarily fully trust and 2) not have to go full perfect 100% of the time opsec to do it" and gratefully signed up. Anecdotally, this class of IP harvesting has gotten a lot rarer, largely thanks to Discord.
Most of these people, despite often having some degree of technical sophistication, don't care about Discord privacy-invasion or censorship because it isn't in their threat model - they've got more immediate fish to fry. Should it be? It's easy for us to sit in our IRC ivory towers and say so, but I think having a communication service where your privacy isn't contingent on the local admin whose court you end up in being a good person is an obvious advantage of centralization.
Of course, Discord isn't immune to issues of compromise (when you're the central point of a juicy system, attackers will try, as shown by Twitter's Saudi subversion thing and various other examples), but compromising a Discord employee is a higher bar than compromising the person running a 500-person IRC server out of their basement. Likewise, many of these sort of privacy issues in decentralized and/or federated services have technical mitigations, but not requiring technical mitigations is an advantage in and of itself.
Anyway, that's how Discord was plausibly a security upgrade for many people I knew who switched to it from IRC/Mumble. I dislike closed, proprietary communication protocols as much as the next person, but I think we've got to treat server subversion/local corruption as a higher class of problem than it usually is treated as, though that gets into a lot of hard problems all the way down the stack.