I just completed a quick-and-dirty test of my proposal, and… nah… a naive implementation doesn’t give very good results. For starters, the linetrace definitely needs more than one sample in a few different directions radiating from the source to get some sort of “average hit location”, and then you get all sorts of complications to consider. While the idea works in principle, as a way to get an approximation of bounce light around the main target of the light, it would take quite a serious effort to get something that looks natural, particularly when moving the flashlight around quickly in any sort of complex interior.
If you’re not interested in investing a lot of time, I’d advice against trying my suggestion.