Looking at the current support geometry, it may be fair to conclude that it’s a bit clunky, and likely wastes material. One possible remedy may be to think about the geometry to support the model during build in terms of a ‘closest packing of spheres.’
See: https://en.wikipedia.org/wiki/Close-packing_of_spheres for a good visual.
In detail, if the volume that defines the domain of the support system were modeled as a series of closest packed spheres, such that if a sphere of R1 were used, there would exist spaces between the spheres everywhere except at their tangencies. If these spaces were instead solid material, and the spherical forms void, the resulting structure would resemble an open cell foam.
By varying the sphere diameter while maintaining their coordinate centers, (in this case R1+n) such that their shared tangency ‘points’ became circular plane definitions, one could reduce the amount of space between the spherical surfaces, and thus reduce the amount of material used, as well as ruducing the resultant strength of the structure. Similarly, in an R1-n scenario, one would increase the amount of material, and commensurately it’s strength (essentially making a closed cell foam).
Now, by varying the diameters of the spheres as one gets closer to the supported surfaces, (in ‘octaves’, as it were) a more consistent support across the surface might be acheived. e.g. larger spheres at the base, decrementing to smaller spheres at or near the surface to be supported. Possibly, the foam would actually terminate just below the surface it is intended to support, and triangulated tendrils (e.g. a tetrahedral frame) would extend up to the surface for a small circular point of contact. The density of these points and their diameters would be configurable, and dynamically changable.
intuitively this would be very strong, use much less material, and leave a much smaller mark on the surface being supported.
Anyway, food for thought.