I have it running for one client (as opposed to double-digits of prosody/XMPP and synapse/Matrix) and for their very limited use case, it works. I haven’t attempted converting anyone else over there and I think there are two main concerns for wider adoption.

uno, I happen to know a bit about the infra it’s running on, basically a dovecot/postfix/postgresql stack for the majority of cases. that thing, although battle tested and widely documented and supported, isn’t without its quirks for the intended use as an instant messenger. there are issues with long-running those services and amateur-hosting those things is a challenge. e.g. each “message” you send is essentially a separate e-mail, and so is the reply, and so on (completely obscured for the enduser). so, in 15 minutes of “chatting”, you’re creating entries in a mailbox that would take months to fill.

dos, the client apps have bad UX and are downright spartan compared to the eye-candy-rich counterparts like Telegram or iMessage, thus hampering adoption from unmotivated users; the users had to be forced (as cajoling didn’t work) to use the thing for its intended purpose and not take their correspondence to accustomed message platforms.

so, both those things can be worked around but I’d caution anyone to not jump headfirst without testing things out thoroughly.