What actually happens in the seconds after you press start
Most sites describe this as if it were instant. It is not — there are a handful of separate things happening, and almost every "it did not work" is one identifiable step failing rather than the whole thing being broken.
First, the mode decides the queue
Video, voice and text are not three settings on one queue — they are three separate queues. Picking one determines who you can be matched with, because the other person picked the same one.
Then your device is asked, not before
Camera and microphone permission is requested when a video match is connecting, not when the page loads. That ordering is deliberate: there is no reason to ask for your camera while you are still reading.
If you have previously denied it, nothing will be asked and nothing will open — see the permissions post for how to reverse that.
Then a connection is established, and only then does it count
A match is not treated as a real conversation until the first video frame actually arrives. That matters for what you are charged: a connection that never produced a picture is not a call you had.
Skipping ends this one and starts the next
One tap closes the current conversation and puts you back into the same queue for a new person. You do not go back to the start screen and you do not re-pick your mode.