Looking at the time stamps (the mtime column), I'm guessing the minimum 'tick' is 10ms (assuming the time from the glitch to the last data point is about 10 minutes).
So looking at that -14,144 @ time 1427189420 and altitude 37900 @ 1427189470 the aircraft looks have lost ~100m in 5 seconds. However, the error bars on that are somewhat unknown, at least to me, although from other entries, ~25m seems to be the resolution
So the glitch may just be an extrapolation of a short duration event to the projected loss over a minute.