Without being able to identify the issue easily from the video I'll have to take a stab in the dark and say you are experiencing something as a result of floating point precision. The pixels can be only rendered on screen in increments.
You could try and see if some form of supersampling (MSAA) might help alleviate it or you could just try updating your motion vector as a integer cast value, making sure that the effects of multiplication by delta factor in whole numbers in the vector components. That is, the motion_offset only has whole numbers added to it.