This requires some sort of an optimization and I think it could be solved with Kangaroo.
I’m using EleFront to reference and bake a sphere center point. Every time a sphere center point is baked and the solver reset, the solution should get more accurate and the distance outputs of the Brep CP component should converge.
One thing to be aware of is that, unlike CircleTanTan, there can be infinite solutions to any combination of two surfaces and a single point because you are introducing the additional dimension.
For example, with these two surfaces and a point at the origin: