No, still not clear. Let me describe my experience. I slew to a star at southern meridian. I accomplish GOTO solving and slewing to target. After that I choose Polar alignment and Measure Az error. It solves, slew 30 min away, solves again and give me the Az error. I press Correct Az error and it again slew, somewhere in between current and previous positions. What star am I supposed to center?

After it calculates the error it should slew back to the original star you centered iirc. You can recenter this star then press correct error and it will slew off to one side and then you use the az adjustment on your wedge to recenter that star. It's a little trial/error at first if the slew causes the star to leave your CCD fov though.

In fact, it might be useful if you could limit how far the scope can be slewed so that the star is always within the field of view and hit a button each time you recenter to slew the scope again and again. So rather than one big adjustment to recenter, you might make several. That way it's obvious which star and also which direction to turn from the very beginning. If this feature was added, it'd be useful to have it easy to toggle on/off though as once you get used to it, you may prefer to do a single big adjustment and recheck.

The way I did it is look in kstars at which way the scope has been slewed, then I adjusted the az a little bit and did a goto plate solve. This showed which direction the correction had moved the scope, the correct direction would cause the scope to move towards the star you originally centered. Then I recentered the star and did the AZ error measure/correct procedure again, this time knowing which way I needed to adjust the mount.

That worked perfectly for alt and once the polar error was reasonably small, the star you needed to center didn't move too far out of the fov and quickly came back into view with a small adjustment. For Az though, I experienced the problem mentioned in an earlier post, moving the scope in a way that recentered the star made the Az error worse. If that happens for you too, please let me know and if it does, all I can suggest is to use the Az measure but not the az correct, you'll have to tweak the az wedge adjuster, manually recenter and re-measure and itterate in that case.

It didn't slew back after error calculation Something wrong on my side? Ekos bug?
What about the sync after polar alignment? If it slews off the stars used for correction, it means sync is lost in the end. Am I right?