I'll suggest it again, just to be sure someone reads it (I know you guys are just experimenting for now, and I know SR is not a new thing): it would be great to put SR in HDRMerge.
Most images need 3 different expositions (-2/0/+2) for near full dynamic range. The SR you presented uses 6 images, so we would need 18 photos to start working. Not practical for many applications, but it would be very useful for landscape or low-light static scenes, for example. Imagine you get a low-end camera and produce images with the quality of a hasselblad. No noise, +16 f-stop DR, no color artifacts, no aliasing. No need for denoising or deconvolution. It would also be good for scientific community, to generate more close to the real images.
HDRMerge already has a align algorithm built in. The SR implementation you are working needs a debayered image to work? Because HDRMerge does everything on DNG, that's the ideal.