I know the post is a bit old but this odd behavior still exists. I’m editing RED footage and audio was recorded into the camera. (Same issues with other cameras as well) I have to sync the audio to match how the camera records it which is 1-2 frames before the visual slate hits. So in other words, looking/listening at the video footage, I place a marker and in mark where I hear the slate which is before the slate impacts. Plural eyes match by audio’s visual waveforms and should sync perfectly.
I don’t understand the science behind that as I was always trained to sync right where the slate closed. It doesn’t seem to be the case anymore.