For your first point, the stage calibration routine measures how many steps there are per pixel on the image, which makes more sense of the movement in steps. It is not a distance but at least relates to something physical.
The steps per pixel from stage calibration, together with the stage geometry would give an uncalibrated scale estimate of pixels per mm on the image. This is potentially very useful and possible. Would you be able to put them into an Issue in the project Gitlab? Issues · OpenFlexure / openflexure-microscope-server · GitLab. A proper calibration routine would be better, but needs a calibration artefact. I have tried a blank CD-R, which is a reasonably common item with a reasonably well defined spacing. It is rather hard to image well enough unfortunately.