After doing a polar align, I notice that the pointing model that is derived from that happens to be degenerate in (like) 60% of the time. Ok, what do I mean by that:

The PA process plate-solves three pictures around Polaris, that means the orientation of the celestial sphere in relation to mount coordinates is determined by three points, that all are next to each other. This means that the accuracy of the solution can diminish fast away from the pole and lead to large pointing errors.

Behaviour that I observed is that the first slew after PA is way off, then taking a new plate-solve (sync) does not fix it. You can even experience that the scope slews to below horizon after that solve or more than 120° away to your target. I even on one occasion, directions where inverted for DEC, but not for RA (or vice versa). My „fix“ so far has been to clear the model or even to power cycle the mount and restart Kstars/Ekos. Then do a plate-solve (sync) in the vicinity of my target, which then usually gives dead on slew-to-targets.

Questions on understanding:
1) Are the PA solves used to built the pointing model or not? (My opinion: should not be used.)
2) What pointing model is there after PA?
3) Where is the pointing model (using EQmod with a AZ-EQ6 GT Pro)? Ekos? Driver/EQmod? In the mount?
4) What exactly happens on „Clear Model“ in mount tab?
5) Can I see the transformation function(s) somewhere? 

Question on how to mitigate:
7) What should the exact procedure be, to get an accurate pointing model? Please assume mobile use (PA necessary each time).
8) Would pointing, ca 5° off of Polaris during PA help to avoid that? 
9) If the pointing model is degenerate (as above), is there a way to avoid restarting Kstars/Ekos?

Thanks in advance, Jens (Grimaldi)