cantiDiscord
First, Discord requires a few settings to be checked off for optimal performance, we'll try making it more obvious, but stuff like automatic gain control and hardware acceleration should probably be played with if it causes issues. If someone is extremely quiet, tell them to reset voice settings first, as internal volume levels could be completely messed up.
Second, Discord works extremely well for us because the current chatroom system in FACEIT is limiting, we'd have to have a Discord Guild for team communication AND team to user communication. We've built upon Discord voice channels because it works well and having a *third requirement* would be non-negligible and sub-optimal, setup couldn't be easily verified, etc.
Third, I'll be working on a Mumble -> Discord voice bridge that will be compatible with TF2PL *whenever I have the time to do so*, it'll end up being MIT-licensed open-source software, if you would like to contribute or start early, please contact me, but please read about these subjects before: https://mumble-protocol.readthedocs.io, https://discordapp.com/developers/docs/topics/voice-connections, and the codec OPUS. I insist on *whenever I have the time to do so*, I've committed to making HUBAR.it (and TF2PL) a day job, my priorities are in releasing the product first to make a living, then whichever better stats from demos/Mumble->Discord bridge has most interest shown.