I had similar ideas going through my head on what the extra resolution could be used for. So imagine cameras capturing at 8 Megapixel and up (so bigger then 4K) and using the wiggle room to shift the frame around using the inverse gyro to output smooth footage. So for every target resolution cameras would actually would need to capture more the necessary. To prevent storage explosion, at least on a consumer-level, the processing could be done in camera memory. Discharging unused data after stabilization is applied. In the professional sector one would expect to be able to fix glitches on stabilization in post … so retaining all the data would need to be an option.
Definition from Wikipedia –