The last image from DART only includes about the top ⅒ of the image, but that portion of the image is in full quality. That suggests that the image encoding/compression is mostly a raster encoding. Off hand this seems to me like a rather odd choice. It prioritizes information that isn't very useful over data that would be more useful.
Does anyone know (or have any informed speculation about) why the images were transmitted the way they were?
The first alternative that comes to mind would have been some sort of global/progressive encoding (e.g. something as simple "progressive JPEG") that first sends low quality image data for the full field of view followed by progressively more detail. The potential advantage of that would be that, regardless of where the transmission is cut off, you get some information from the whole field of view at the expense of not getting really high quality.
The second alternative that comes to mind is even simpler: just re-order the pixels to transmit the center of the image first. The point of impact is likely the portion of the last image that is most interesting and the encoding used has only about a 50:50 chance of showing that.