Do note that this was prepped with a ton of extra contrast, much beyond usable situations, but makes it easier to display what I was seeing before any contrast was added.
I do want to say your ISO 400 f/5.6 clip seems wonky, the brightness is uneven, perhaps a mistake in the test. So maybe this 1600/800 discrepancy was also a mistake. Please try again and see if you can verify what I'm seeing here.
Maybe the added noise is dithering the noise pattern a little? Anyone else?