I think this problem left and came back a couple times.. but here is an example track.

Listen for row 42 when they lose sync. They should stay in sync if I am not mistaken.

the fix that was added back then caused a lot of other problems (such as >>3632, if memory serves me correctly)