Hmm, still don't see anything untoward in the data.
In case you are curious, attached is a rough graph of the speed the mount is going based on differentiating the encoder values from the log files wrt to the timestamps in the log. The wigglyness is mostly due to the way I processed the data (the timestamps are not equally spaced so I had to interpolate and resample).
Actually this is the first time I've plotted the data like this, and it shows how my approximation of linear acc/dec is sort of working, but could do with some more work at the low speeds to get rid of the weird tails.