[aerohead]

I did a rudimentary 'road test' of my tennis racquet. It's an adult, HEAD, Liquidmetal, Supreme. It's a light metal hoop / strut / handle, strung with 1mm-diameter Nylon. The mesh opening dimensions vary, depending upon position within the inner boundary of the distended hoop.
The 'most dense' pathways are near-center, creating a rectangular opening, 13mm by 10mm.
Due to flow separation, the wake, downstream of the 1mm string is 1.244mm, resulting in a vena-contracta-esque narrowing of the airway to 12.51mm by 9.51mm.
--------------------------------------------------------------------------------------
1) Just 'swinging' the racquet produces an audible acoustic signature.
2) At 35-mph, it begins to 'sing' appreciably.
3) At 60-mph, it's 'roaring' significantly, and my wrist tired rapidly, attempting to steady the racquet in the oncoming airstream.
4) I've no means to parse out, and properly attribute the sound contribution and associated drag between the hoop and string grid.
5) I'm unwilling to invest the time to 'streamline' the hoop, and re-test.
------------------------------------------------------------------------------------
I suspect that there'd be little trouble coming to a consensus that, the noise was an artifact of separation, and any use 'upstream' of a road vehicle's body , or components would be unacceptable.
-------------------------------------------------------------------------------------
Onset flow to side mirrors is typically 3-dimensional, making alignment of 'ALL' of the grid, normal to the flow downstream of a mirror problematic.
--------------------------------------------------------------------------------------
I revisited Syed R. Ahmed's flow traverse research reported between 1979 and 1984. The 'densest' concentration of tuft intersections, in a half-body- width grid, was for the fastback body, with a Cartesian grid of eleven vertical columns, and thirty-two horizontal columns.
Based on the metrics he used for grid sizing, for a 2018 Nissan Leaf, imaging would require a grid of 73.6-inches height, by 42.3-inches width, with 352-tufts, spaced at around 3.5-inches, by 2.25-inches. Perhaps this would be a reasonable tuft population density.
--------------------------------------------------------------------------------------
'Streamlining' the grid elements would introduce a 2-D flow bias within a 3-D flow environment. One might get some of the grid properly addressed with respect to some of the oncoming flow, while the rest of the flow would be unaligned, introducing transverse contamination of the flow field of interest.
-------------------------------------------------------------------------------------
Tufting the side mirror itself would preserve fidelity of the actual flow.
-------------------------------------------------------------------------------------
Since Hucho et al. published their shape optimization research in 1976, and new car production cycles back then ran about 48-months, it's hard to imagine a new production car manufactured after 1980, which would demonstrate significant flow issues for their forebody.
' Saturation' leading edge radii relationships were known by this time, and by 1969, the world community of auto designers were in possession of shape 'recipes' which would produce Cd 0.245, as of 1969.
------------------------------------------------------------------------------------
It seems to me that if everything published by 1976 is addressed, then lower drag will just come down to body elongation, along known streamlined contours/ silhouettes, also reported by 1969, if we ignore those known for over a hundred years now.