Pixel shift isn't that useful for simple physical reason: shifted pixels overlap. With microlences on top light from the whole frame is collected. You can imagine for simplicity 1D array of segments. First pixel covers segments 0 to 7, second 8-15, and so on. Pixel shift will make them cover 4-11, 12-19, 20-17,... So, you get 2x pixel, but don't actually get 2x resolution. Because they overlap. To get 2x you need 0-3, 4-7, 8-11. It can be produced only with some accuracy by algorithm or AI.

Conclusion is that supershift is better then interpolation, but not a lossless 2x magnifier. However, it should reproduce colors better. Because for each pixel we can gen 2 or 3 RAW colors instead of one in normal shot.