Depends how much you do/don't want to customise Adapt. If you need to be using Adapt out-of-the-box then using two media components side-by-side is probably the best option and yes, making them mandatory for completion wouldn't really work.

If customising Adapt is an option then I'd look to add some kind of accessibility settings feature that the user could be presented with at the start of the course that would allow them to select things like audio description and you could customise the component to switch out video files depending on what setting had been chosen.

Do you know if any screenreaders support it?

No I don't, sorry

I think in the case of most elearning content, screenreader users are generally very well served by a transcript of the video as this can be parsed and read back by screenreader software at high speed rather than being constrained by the speed of the video. It really depends on the video content; i.e. 'talking head video'. But I appreciate that if you have a client who insists on sticking absolutely to the line of the WCAG then any attempt to take a pragmatic line is doomed to failure - particularly as they are guidelines and not rules and so (deliberately) open to interpretation...