I cleared the alignments in the Alignment tab of the device manager and in the successive plate-solve done during the polar alignment.
Moving the mount "with the virtual pad" we point a star some degrees out of the pole zone. The astrometry systematically fail.
The point is that it only recognize the polar area in astrometry but was lost everywhere else.
When trying to point an identified star it start to:an unexpected direction.
One hypothesis there is some info that wasn't really deleted.
Another point is that the single board computer on which Kstars, Ekos and Indi were running has been connected to an AZ-EQ6 before and worked well.
Another hypothesis there might be common parameters in the configuation file of the driver are shared by several mounts and can pollute the way that, they, work.
Another hypothesis could be that the time of the mount and the time of the SBC are not synchronized. If that occurs the question is why and how ?