While watching the SES-10 flight, starting a little after T+1:30, I noticed that the direction of the first-stage exhaust plume started to diverge from the alignment of the rocket body.
The divergence angle increased continuously until the plume became too tenuous to clearly determine the direction it was pointing.
After the 2-min mark, the plume appears something like 30 degrees off-line (quite foreshortened, so the actual divergence angle must be smaller).
I thought briefly that it might be deflecting to correct for an engine shutdown, but from 2:02 to 2:14, the 8 separate plumes from outer engines are distinctly visible and angled in the expected directions.
I would expect any deliberate angle-of-attack on ascent to be quite small, and that a gimbaling of the engines that substantial would turn the stack very rapidly. At that speed and altitude, dynamic pressure should still be substantial (perhaps 7-8kPa?).
What explains this observation? Is this just the effect of a strong crosswind, either pushing the plume aside, or forcing the engines to gimbal to correct for the rocket being pushed off trajectory?
Here's a more complete set of images in the sequence.