I replaced my old microswitch from the Klicky Probe and Voila! convergence is more than perfect. So finally it was a bad (damaged) probe.
OK, Thanks all for the contribution. I’d like to continue the conversation about QGL and Z-Tilt in the Feature Section, if there is interest in. What I want to do is to change back from “NG Klicky Probe” to normal “Klicky Probe”, if I got some spare time (never).
EDIT: This is the repeatability of the new micro switch (again: 0,10€/part). Standard deviation is 0,004mm which should be good enough for my case (Hey REXT3D: Should I try the 0,005 again? )
09:20:25 // probe at 300.000,275.000 is z=7.464237
09:20:27 // probe at 300.000,275.000 is z=7.467987
09:20:28 // probe at 300.000,275.000 is z=7.467987
09:20:29 // probe at 300.000,275.000 is z=7.471737
09:20:30 // probe at 300.000,275.000 is z=7.475487
09:20:31 // probe at 300.000,275.000 is z=7.474237
09:20:32 // probe at 300.000,275.000 is z=7.474237
09:20:34 // probe at 300.000,275.000 is z=7.476737