Typically, the central server does not transcode. Participants simulcast a few bitrates, and the central server forwards to each other participant the sub-stream with the appropriate bitrate for the bandwidth capacity of that participant. This is compatible with E2E encryption, by individually encrypting each sub-stream. Participants can share a session key that is unknown to the central server.

FaceTime supports group calls and claims E2E encryption for them. WhatsApp does too, I believe? I'm not sure how many participants you can have.