Surface reflectivity can be an issue, that's true. And the numbers aren't as easily compared as the simple geometry of casting shadows. And I assume that there must be large differences between different module types?

The big difference is the kind of glass and coatings used, this can make a few % difference at the same angle of incidence. Typically the yield below a 45 degree mismatch is the maximum panel yield x cosine(angle of incidence) x transmissions (anywhere from 90 to 95% depending on coating and base material). Beyond 45 degree mismatch there is a sharp drop off.

Certainly a topic where it's impossible to judge without knowing actual curves. Even looking at actual panels (without measuring output) you'd be prone to misjudge because some that look promisingly low in terms of specular reflection might pay for it with higher diffuse which won't notice as long as it's not particularly bright.

That leaves armchair-level pondering to theoreticals: is the perfect "any angle" surface even physically possible or are there physical limits that only leave tradeoffs?

And another aspect I don't know: do the cells mind? Does that "punch an electron" principle still work out when photons come in almost parallel to the electrodes? (in practical terms: if we compensated the lower effective cross section and reflectivity with higher light intensity, would we still get the same voltage?) When I was a kid I had a miniature cell in an experiment kit that had a deep plastic top layer with angled prism structures under the surface almost like those of a retro-reflector and I always wondered why it would have that. Was it because the cell could only collect from photons coming in almost perpendicular, and the job of the plastic structure was to make sure that there would always be some photons making the angle threshold (certainly by sacrificing a huge amount of power in the well-aimed case)

This all almost sounds like an argument for heliostats (which get the best yield per module surface, but the worst yield per acre), and which would be a perfect match for agrivoltaics. But that's an approach that hinges entirely on the cost and (far more importantly, at scale) resource use of the mechanical structures required. Which is a mechanical engineering problem that you'd have no problem explaining to a victorian era engineer. Genius "inventors" to the rescue?

I have 26 panels set up like that of two different types and the figures pretty much match that formula.

I've used two heliostats in the past (see other comment in this thread) and on a $/Watt basis you're better off putting them flat, regardless of the perceived advantages, after all is said and done you will have more power, be out less money and have a more reliable system.

Now if only the sun would shine...

That possible heliostat future I keep bringing up wouldn't be about $/Watt (that was an argument for helostat back when modules were much more expensive, but certainly isn't the case anyone), but about finding a sweet spot in agrivoltaics between energy and nutrient harvest: Watts per impact on farming. Or more specifically, electricity dollars per impact on farming. Because when there is a large installed base of photovoltaics, watts at noon won't be worth much compared to watts at deeper sun hours. And for the same amount of off-noon watts, a heliostat setup would cause less shading to the plants below than any other setup. And as a bonus benefit, the east/west spacing inherent to heliostat installations would give a nice distribution of shading, at least nicer than with fixed east/west lines. I believe that the advantages are quite clear, but of course only if the mechanical parts can be cheap enough (in terms of resource use). Flat north/south lines that only pivot east/west would surely be a sweet spot, up to a certain latitude.

The amount of land required to supply ample energy to some random region (bar Switzerland) is usually a small fraction of what is dedicated to agriculture / barren land.