I think that after having that dialog a few times, most users would disable it, because it's not really user friendly to actually have to press a button each time a video has ended. At least, if I'm correct in assuming that most users run the GUI in the highest possible refresh rate (for me that's 60Hz), and videos in their original refresh rate (like 24Hz/25Hz/etc.)? In normal use, you wouldn't want the GUI to interrupt you when a movie has ended. Especially not when it would be useful - i.e. when you're watching several videos after another.
I do see OP's point that it's irritating to wait for refresh rate changes, and I'm just using a normal TV that only takes one or two seconds.
I would guess for the OP (and others) it's not a problem when the refresh rate only changes after watching a full movie, because a refresh rate change would then just occur like once or twice in an evening of watching video. It's, like OP says, only a problem with several changes after each other, i.e. when watching short clips (movie trailers, YouTube videos, videos in picture galleries, etc.)
Could a solution be to add an option to change refresh rate only when the video that's being loaded is longer than a certain pre-set amount of time? That would mean that the TV or projector would change to 24Hz when you're watching a full movie, but doesn't change when you just want to watch a 2 minute trailer, and then return back to the GUI. (I guess live streaming like in PVR should always be assumed to take longer than the pre-set duration.)
I realise that this is the opposite to the OP's proposed solution of not-changing-back to the GUI refresh rate, but I think the user experience is better this way. At least, for me it's more detracting to have the GUI run at 24Hz (or something else that's slow) than it's to see a short Youtube video in 60Hz instead of 24/25Hz. For me personally it's relevant because I often use Kodi to run picture slideshows with mixed context of pictures and videos. Kodi changes refresh rate before and after each video (normally 24Hz), but keeping the screen locked at 24Hz makes a photo slideshow pretty horrible.
Now that I've typed this, I see advantages to both proposed solutions.