Hmm, I'd say this is the 'rain' I had been talking of. You can see that the 'stripes' are in RA direction.
The pattern is made from so-called FPN, fixed pattern noise. Think of it similar as hot pixels, just that their deviation is much less. So you do not see it in a single exposure, as they are randomly distributed. But if you start displacing it in only one direction those deviating pixels will start to form the lines you see there. This is why dithering isn't really effective in RA-only guiding.
If I understand correct, for the first image dithering failed. Would really be important to know if it only claims it failed, or if the dither really did not happen. As you say it works with Stellarmate - is that indeed also with SEP and multi-star guiding?