Extensions on Magellan Megacam images

I am working on astrometric registration of Magellan Megacam images. My first task was to read the multi-extension FITS files. The image files I have possess 36 extensions, each corresponding to one of Megacam's 36 detectors. The order of the extensions follows four columns of nine detectors each, starting at the southwest corner of the mosaic. Thus extension 2 is for the detector immediately north of extension 1, and so on up through extension 9, which is the ninth detector of the first column and hence is at the north edge of the mosaic. Then extension 10 is for the first detector of the second column, and is immediately to the east of extension 1. This is a very sensible way of arranging the detectors and extensions.

One thing that confused me at first is that although the third and fourth columns have the detectors ordered exactly as you would expect based on the above, each detector image in these columns is upside-down relative to the corresponding images in the first and second columns. One can understand how this might come about due to spatial constraints on the positioning of readout electronics.