Number 4 in your list is the description of the star tracker going out of synchronization so that the computer thinks it is receiving pixel (1,1) but is actually receiving pixel (1,3) or pixel (8, 7).  There is no signal being sent to indicate the borders of the 8x9 pixel frame.
 
When this happens the star will, in effect, be moved by  one, two or three pixels horizontally. If there is a larger de synchronization, it might even be shifted up or down by a row.

The source of this problem is noise on the toggle line which is interpreted as the signal of a pixel being loaded onto the DIO card.  The source of the noise may be electrostatic discharge from rubbing by curtains or something
touching the data wires.

My normal test for this condition is to turn the dark fielding off and to put some light on the CCD's to ensure that the ninth column remains darker than the rest of the chip.  The appearance of a bright vertical bands
corresponds to the dark field being applied to the wrong pixels.