I knew a Japanese architect who insisted he made these sorts of uncanny breaks deliberately. In my courses, they come about by accident, when the shaper’s eye is fooled - and I will usually leave them in.
My old friend Ed Connor, who was the first to do detailed mapping of greens in the 80s, told me that the Golden Age architects and builders’ eyes were frequently fooled, too. He said that while Rosd and Tillinghast normally put 4-4% tilt in their greens, he had notice they erred toward 5-6% when playing into an upslope, and sometimes wound up too flat when the green site fell away to the back. It’s possible some of that was deliberate, for visibility of a green above you, but Ed didn’t think so.