I don't fully understand the intended working of the sync TAB in EKOS, but there are three choises: "sync to target", "slew to target" and "do nothing".
I think "sync to target" should tell the driver where the scope is actually pointing, and the driver should update it's position so that the crosshair points to the newly learnt position and keep tracking that position.
In my understanding the "slew to target" should do the same, but affter that it should issue a goto to the original target.

The "AltAzSimple" driver behaves something like that. But the APIMount driver does not. After a "sync to target", it updates it's internal knowledge about the scope actual position but leaves the tracking target at the original position so the tracking code feels a tracking error and corrects. Effectively getting the same result as the "slew to target".
I have come to the conlusion that the difference lies in that the APIMount driver keeps the tracking target in Ra/Dec coordinates but the Simple driver keeps it in AltAz coordinates. At sync operation the Ra/Dec coordinates of the scope gets updated, but the AltAz coordinates remains the same (the sync operation updates the Alt/Az -> Ra/Dec mapping), so the driver keeps tracking the same point.

To answer to your question: sometimes plate solving fails for me without any real good reason (there seems to be enough stars on the taken image, the actual target is close enough, no good reason), sometimes later (maybe after moving the scope a bit) the plate solve just succeds.Also sometimes I got the error (after succesfull solve) "sync failed". I have no idea about the reason. Seems random to me, but maybe I just dont see something.

I have no real success with guiding. I have succeeded the calibration a few times (most often than not, just calibration fails). But after start tracking it tracks very rudely, much worse than without tracking. I have given a try to PHD but no success. After calibration it has strated tracking but it just diverged instead of tracking.
So I do multi-hundreds of short (10 sec or so) expos by splitting it apart, say 50 picture at a time, the re-center the target and shoot an other 50. When it drifts so much that the target drifts away in 50 shots, the individual pictures will be bad qualtiy (due to poor tracking) anyway. And as you said, image rotation is inevitable for AltAz mount so one must have enough headroom on picture framing anyways.

Read More...