Thank you for the interesting photos, Steve.
Please note that whenever you do flow visualization on the surface of an aerodynamic body, you are seeing what the boundary layer is doing.
The air 1" away from the surface may be doing something quite different.
The boundary layer is moving slower than freestream, has less total pressure, and consequently can be turned at steeper angles by pressure gradients than the air away from the boundary layer.
This is a known issue with oil streak testing of wind tunnel models, for example. If there is a small local flow separation, the oil drops may move backwards, sideways, or in lazy circles.
I was once involved in oil drop testing inside a large commercial turbofan engine fan duct. The oil drops were placed in a rectangular array on the ID of the fan duct, downstream of the fan. It was remarkable where some of the oil drops ended up - some of them crawled radially outward along the trailing edges of the fan outlet guide vanes, and streaked back along the fan duct OD. I"m quite certain that on average the fan duct air doesn't turn over from the ID to the OD of the duct, but the boundary layer flow can in fact migrate around quite a bit...
Best Regards,
-Paragon
Cincinnati, OH