It is interesting to look at the differences in the video:
First, there is a difference in the trajectories – the launch from Dragon CRS-8 was performed on the ISS, and from SES-10 to the geo-transfer orbit. In the latter case, because of the insertion into an intermediate orbit with an 18-minute pause between the first and second inclusion of the second stage, the trajectory turns out to be more shallow, so in the start with SES-10 the speed increases faster and the height slower.
Secondly, the shutdown of the engines of the first stage in the case of SES-10 occurred about four seconds later. Most likely, this is due to the fact that the over-caliber fairing experiences increased aerodynamic loads, and at the maximum speed head (Max-Q), the thrust of the first-stage engines decreases more strongly or for a longer time. UPD : in the comments in LJ found an analysis with Reddit, and on it, on the contrary, missions with the Dragon, including CRS-8, reduce traction for a longer time, and according to the scheme it turns out that the step with SES-10 simply ran longer, and more fuel remained in the tanks at the start of the CRS-8.
Thirdly, from a flat trajectory it would be logical to expect an earlier landing, but here in the case of SES-10, the level of the village is about ten seconds later than the CRS-8. Here everything rests on concrete inhibition algorithms, and, without an exact knowledge of the flight parameters, it is difficult to say why it happened.