Thanks for doing this comp. I didn't expect this big a difference. Are you sure that the orientation of the mic capsules didn't shift while you were switching cables? Because A and D sound dramatically different than B and C, far more so than a preamp switch should. One of the pairs sounds very rolled off in the treble. Is it possible your mics shifted, even slightly, or you inadvertently had some filtering engaged on one of the recorders?
^^^^ I agree that it certainly sounds as if there is something in the test methodology skewing the results. The 788 is a legendary machine, for good reason and I would expect to hear "some" difference.
Is there a possibility that one or more of the following was not equal?: High pass engaged/different frequency selection; Limiter engaged/level setting; Preamp gain difference (you noted that you did do some level processing in post?).
I am seeing differences in both level and stereo width between both test groups when monitoring via RME DigiCheck....
As part of my post retirement gear purge I recently did a mic shootout of all my mixers and recorders to see what I could/should justify keeping for personal use. Products included from Yamaha, Sony, Presonus, A&H, Mackie and Sound Devices (MixPre, MixPre-d, USBPre 2 (closest pre topology to 788), 552 and MixPre6ii. They all sounded very slightly different, as expected, and I did come to the conclusion to keep 2, maybe 3 (all winners were SD machines). The differences were slight enough that the final determination was based on channel count, portability and integration to a projected workflow for future projects.
I certainly don't question your sincerity and integrity in doing this test, but your results vary much more than has been my experience.