Because the camera records RAW, and to do downscaling, it would have to first convert it to RGB, tripling the amount of data to record if left at the original resolution.
RAW contains a mosaic pattern of photo site information filtered green, red, and blue so you can't just drop every other pixel in a row to create a RAW file that is half the resolution. It has to be de-bayered into a full color image before it can be properly downscaled. That's a lot of processing and more data to record unless you are dropping way down in size like from 5K to HD. But 5K RGB is basically three 5K files for one color image so is 3X the size of a 5K RAW file.
The Sony F65 also records 16-bit raw information and leaves the conversion to 4K RGB later.
Sure, you can do a lot of downsampling in post but generally you are working with a bigger computer than the one inside a camera and you don't always have to do it in real-time.



