IMHO the alleged drag losses on the upwind side of the Savonius are overrated - especially with blade profiles like those of the Sandia design. The curvature of the blade provides lift for a significant part of its upwind travel - until the trailing edge is nearly pointed downwind. By that point the front of the blade is directing significant amounts of additional air into the the downwind blade, giving it a boost that compensates for much of the upwind blade's drag.
(If you look at the sandia savonius efficiency numbers and compare them to that of a darrieus (or the corrected version of that popular efficiency vs. TSR graph) you'll find that the sandia savonius is approaching the efficiency of a darrieus and is significantly above the calculations based on losing a third of the power to upwind drag.)
While a shroud might do a better job of eliminting the drag, it's a complication and if not properly designed it may also lose more by interfering with the upwind blade's direction of wind into the downwind blade than it gains by shadowing the upwind blade.
Not to say you can't do better using a shroud. But you'll have to redesign the rotor in tandem, discarding the optimization research results for unshrouded rotors.
I'm inclined to believe that, engineering-wise, it's better to stick with the KISS principle and compensate for any inefficiencies in an unshrouded Savonius rotor by just making it a tad bigger. (Moderate differences in efficiency aren't particularly significant when the fuel is free.)
