Unless swap-move or overwrite-only upgrade strategies are used, the automatic tests executed in the simulator are configured to use two images of the same size for testing the upgrade and revert processes.

Indeed, the size of the images seems to be defined by this routine:

I agree it is a good thing to test the upgrade and revert processes with the largest possible images but I also think it important to that with two images having different sizes, since equally-sized images is very specific and uncommon scenario and testing only that scenario doesn't enable to detect issues such the one described by #2109, occurring only when the two images have not exactly the same size.