Cool cool cool. I expect that they are measuring the distance between the ears and applying frequency-specific time and level delays from that. They also appear to be scanning the pinna; can they interpolate semi-appropriate pinna filters from that?? That seems much harder to me, but video processing can do all sorts of crazy stuff these days so, sure? Based on some random user comments on reddit, I am guessing that they are not getting as accurate a head related transfer function as they would get from recording audio from directly into your ears. The old example I always think of is the barber shop one, which it looks like Starkey stole and repurpossed for marketting at some point: https://www.youtube.com/watch?v=IUDTlvagjJA (use headphones)
In regards to relevance for individuals with hearing loss, pinna effect cues are pretty high frequency so depending on the hearing loss may not be accessible.
It’s flexible. If you make model pinnas and stick them on your head backwards while smoothing your own pinnas down, everything will sound backwards for a while and then it will flip and seem normal. You can temporarily shift both time and level difference assessments, and therefore localization judgements, by fatiguing one ear with multiple tone exposures. You’re right that the ears tend to follow the eyes. But there’s a combination of flexibility and . . . predetermined architecture. In owls for example, if you plug up one ear at birth you can track how they recover sound localization once that ear is unplugged. Recovery slows the longer you wait and after a certain point (~48 months?) you have passed the critical period and they will not recover sound localization when you unplug the deprived ear. Early experience is protective against later monaural deprivation. I don’t have a reference for how long that critical window or sensitive period is in humans; I’m not sure we have that information yet. (Critical developmental windows and sensitive periods exist for all sorts of things, and are worth reading about if you haven’t. Language is surely one of the most interesting areas.)
I’m tremendously biased, but I think auditory localization is much cooler than visual localizataion. 