Because as soon as you launch, Zoom has to assess the ambience audio and has to prepare sufficient model for noice cancellation profile. By the time you must have observed that Zoom has better noice cancelling AI model than Google or MS team, because of such heavy-lifting. 

Also it would allow you to test the computer audio, these things would require microphone. I wouldn't be worried about Zoom uses my microphone, I will be actually be concerned if the audio is saved and transmitted over internet, which is not the case here.