There is a "Crosshair.Gap Size" property that controls the size of the gap at the intersection of the crosshair planes. The size is interpreted in terms of display pixels.
It would be useful to have a "Crosshair.Gap Size unit" property so that you can define the gap size in millimetres or voxels.
The problem with the pixel unit is that if you work with images and do manual segmentation, you have to know which voxel is selected exactly. However, depending on the zooming and the spacing, the gap can enclose either several voxels or only a partial voxel. Moreover, if you choose a gap size that is comfortable to work with, it might not be good for higher or lower resolution images.
The allowed values could be "px", "vx" and "mm".
If the property is not defined it can fall back to "px" for compatibility. However, "vx" or "mm" would be more useful because the gap would increase and decrease together with the image when zooming in and out. So, I would set the default to one of these for the MITK Display.
When following a more complicated logic, the mapper could check if the input (aka. reference) geometry of the renderer is an image geometry and apply different units for world geometries and image geometries. But in this case there should be separate preferences for the two cases.