It depends on the frequency of the gps updates. If they're regular enough then 1cm accuracy could conceivably show a speed in the opposite direction of travel for one update.
Would need more info to determine whether the data being presented was due to a bug or just acquisition period anomalies.
Ps. If the data is averaged over a few minutes then the starting accuracy of say 1km could out a position 1km ahead of your track, taking a massive chunk out of the average speed for a few minutes.