The old code used standard factor and offset values
to handle coincident surface/line issues. There are known
issue with that approach though. Halos on cones being one
example when surfaces have a factor.

This approach does not use a factor but instead only
offsets. Unlike the previous offsets that were in zbuffer
values, these offsets are in terms of view coordinates.
The advantage being that as the zbuffer is non-linear the
desired offset needs ot take that into account. By using
view coordinates we take the non-linearity out of the
equation. Specifically the offset is in terms of
1/10000 of the diagonal of the actor. So an offset of 100
would push back the zvalues by 1% of the actors bounding
diagonal. Negative values pull zbuffer values closer to the
viewer. The defaults are now 0 for surfaces, -4 for lines,
and -8 for points. It is valuable to have 0 for surfaces
as that means those fragments can use early z rejection
as they are not modifying the z v...