The GPT-4 description isn't too wrong. It's basically an intentional delay where LFS does nothing to allow other processes to do something. Less of an issue on modern multicore systems as there's 2-32 "cores" (cores + hyperthreaded cores) available to carry out tasks.
I'm amazed by how good Chat GPT's answer is. It is exactly correct.
On the other hand, I have also played with this setting in LFS and saw no obvious difference, but I keep it turned off. I also use frame limit instead of vsinc to cap the fps to 120, as it doesn't introduce any controller input lag.
I have a 60Hz monitor, so it looks more fluid if I set 120fps cap instead of 100fps. The physics and ffb engine anyway runs as 100Hz, unless there is a fps drop below 100fps, so all good there.