For mobile devices, it seems user action such as a click/touch is needed in order to play video (or perhaps audio too) - specifically before calling the play() method on a video element is successful.
See for example Can you autoplay HTML5 videos on the iPad?
However, is that for every play() or just the first? i.e. can I have a "start" button which quickly calls play() and pause() on a video element, and then subsequent play() on that video element will not be via user action but rather some other internal logic down the line?