So... for the voicecom/screenshare usecase #xmpp is currently nogo. There is XEPs for 1-to-1 audio/video calls (called jingle) and mesh-varieties for jingle (so, p2p mesh where every participant connects to every other participant). This is absolutely not usable in any way, at least in the groups where I am. Even just pushing eight audio-streams to every participant, where participants are located evenly across whole Europe, UK and MENA... the bandwidth requirements alone are not reasonable and variable latency between every participant would make communicating impossible.
There is (or should be said, was?) https://xmpp.org/extensions/xep-0340.html for Zoom/Jitsi like conferences. That is apparently dead and no one uses it and wouldn't match the Discord like experience anyway with conference conductor.. thing.
Closest attempt for usable voice chat rooms seems to have been https://xmpp.org/extensions/inbox/av_conferences.html which on a surface level sounds like it matches Discordy-y lightweight voicecomms. Sadly Version: 0.0.1 (2024-07-29), so it has never been upgraded to proper XEP.