If Apple wanted to put an engineering team on solving this problem, they could record all the raw sensor data for the video, with the regular 'auto' settings, then, after the clip is recorded, decide what shutter speed, iso, etc to use, and then reprocess that raw data to simulate what that moment in time would have looked like with a different shutter speed.

I''m sure modern neural nets would do a decent job of simulating what a frame taken with one iso/shutter/focus would look like with a slightly different iso/shutter/focus.