With my machine to solve model of 200.000 elements it takes around 35min for 2 step procedure (2 solver runs, GMRES), where Stratton-Chu integration is part of second solver run, so I imagine it takes less than few minutes, but again depends on number of mesh elements that your sphere is composed of.
Making integration sphere smaller-I am not sure. Me and few of my friends are still having some doubts regarding Stratton-Chu formula and its definition. In RF module guide they say that one should put integration sphere in the near-field
of your scatterer. That is maybe correct for dielectrics, but for metallic nanoparticles there are strong evanescent fields, so maybe you have to avoid collecting them into Stratton-Chu. I am not a theoretician so I don't know yet how Stratton-Chu is really defined, and we are trying to clarify it with COMSOL support at the moment.
So far, I am placing integration sphere on the distance few times larger than evanescent field decaying length, and that is easy for simple structure (spheres, etc....)
ANY DISSCUSSION about interpretation on Stratton-Chu is welcome here....