Apparent motion of a target across the detector can be caused by spacecraft-induced motions including drifts, wobble, and jitter.  
Settling drifts on the order of 0.5 arcsec over the course of half an hour after slewing are believed to be due to variable heating and cooling within the spacecraft bus. The resulting small temperature changes cause very slight changes in the alignment between the star trackers and the boresight of the telescope. The temperature changes themselves are believed to be caused by heating of the spacecraft bus when the telescope is observing (or communicating with Earth) at relatively large pitch angles, and radiative cooling when the pitch angle returns to less than +/- 10 degrees. Depending on the amount of time spent at such high pitch angles, the spacecraft may require anywhere from hours to days to reach temperature equilibrium.
On top of the initial settling drifts, there is a roughly constant, long-term drift of 0.34 arcsec per day in the -y-pixel direction. This drift is believed to be due to an inconsistency in the way that velocity aberration corrections are handled by the spacecraft's Command and Data Handling computer (C&DH) and the star trackers. Whereas the C&DH assumes that velocity aberration is constant over the course of each AOR,  the star trackers update the velocity aberration corrections continuously. The result is a constant error signal sent by the star trackers, which the spacecraft constantly attempts to correct for.
Another thermally-induced pointing drift is an oscillating "sawtooth" motion in both x- and y-pixel directions. This is believed to be due to thermo-mechanical expansions and contractions of many of the same spacecraft components as above, motivated here by the periodic, on-off cycling of a battery heater within the spacecraft bus. The amplitude of the sawtooth was ~0.15 arcsec, with a period of about an hour. To reduce the amplitude of the oscillation and move the period away from the typical durations of exoplanet transits, the cycling range was reduced to 1 K on 17 October, 2010. Data taken since that time show a sawtooth amplitude reduced by factor of 2, with a period of 39 +/- 2 minutes.
Jitter has both high frequency (> 50 Hz) and low frequency (~ 0.01 Hz) components, with peak-to-peak amplitudes ranging from 0.03 to 0.1 arcseconds.   Possible causes are many, and may include noise in the star tracker solutions, momentum variations in the reaction wheel assembly harmonically coupled to the spacecraft structure, and even micrometeoroid impacts. High frequency jitter will presumably continue to set a noise floor for bright targets.

Further references: