There are multiple problems with the sound ATM, but what I think you are speaking of is the electronic subtraction of what you just said over the microphone from the returning sound stream (it is because of this that your voice never gets recorded). I’m guessing they do that to reduce feedback or confusion, but there is also an audio delay that can make it last after you have finished talking. Hopefully this will get cleaned up as part of their sound rework, but the best thing you can do ATM is report all sound issues to the developers: