OK, with that position calibration should work, but obviously It doesn't. It seems when EKOS tries to reverse the RA motion after three successful steps, moving back has only effect in one direction.
The first three steps RA increase the x-axis and decrease the y-axis positions:
[2023-02-26T20:44:04.123 CET DEBG ][ org.kde.kstars.ekos.guide] - Iteration # 1 : STAR 146.543 , 246.288
[2023-02-26T20:44:08.505 CET DEBG ][ org.kde.kstars.ekos.guide] - Iteration # 2 : STAR 148.05 , 242.722
[2023-02-26T20:44:12.911 CET DEBG ][ org.kde.kstars.ekos.guide] - Iteration # 3 : STAR 149.442 , 238.707
But when it reverses the direction, the y-axis position still decreases:
[2023-02-26T20:44:17.303 CET DEBG ][ org.kde.kstars.ekos.guide] - Iteration # 4 : STAR 148.179 , 235.289
[2023-02-26T20:44:21.688 CET DEBG ][ org.kde.kstars.ekos.guide] - Iteration # 5 : STAR 146.66 , 231.959
[2023-02-26T20:44:26.097 CET DEBG ][ org.kde.kstars.ekos.guide] - Iteration # 6 : STAR 145.836 , 228.362
...