I'd actually always heard it was from the original lacquer cutting process. On a loud passage the cutter head would have more modulation and this would actually deform the adjacent/previously cut groove from the prior rotation. Sine the dead wax is theoretically dead quiet, the dynamic contrast from the start of the track often makes it's presence known. It's not heard at the end of a track because the next groove hasn't yet been cut and therefore can not be deformed.
I also understand that part of the job of cutting engineer is to actually vary the groove spacing so that really dynamic passages are given more "shoulder" between each groove so that this distortion of a groove cut on the last rotation is minimized on the next.
I also understand that part of the job of cutting engineer is to actually vary the groove spacing so that really dynamic passages are given more "shoulder" between each groove so that this distortion of a groove cut on the last rotation is minimized on the next.