One issue though is the lens entrance pupil, basically the lens front, needs to be about 17mm to give the very minor pop we associate with a 35/2. I haven't seen a cell phone with even 1/10 that area yet. It actually doesn't matter how big the sensor is, I don't think, or at least that affects the result more at the pixel level, but you need to have that glass to get a DOF-based pop of your subject.
Enter computational imaging advancements. Physics can’t be defied, but maybe it can be circumvented completely.