Now I have bought an AF-D lens (nikkor 35 f2.0) and I have same results as with the 35 1.5 AF-S. The AF-S lens had a internal motor and the AF-D lens uses the camera motor. The 50 AF-S 1.8 I have uses its internal motor too. I don't think the 3 motors are damaged.
This corroborates the conclusion that AF is not 100% consistent and is not 100% free of error and as an electro/mechanical control system it has a deadband or error-tolerance. If you add to this low light and little DOF the "errors" will be more evident.