[Zig]

Nice experiment; thank you for letting us all in on it as I find this very engaging.
I reckon it'd be fair to say that in terms of exposure, we all learn to "read" how our sensors behave and expose accordingly, balancing out the noise in the darks and bearing in mind how well our raws are recoverable.
It remains true that highlight detail will mathematically/physically be more "rescuable" and detailed if overexposed, as opposed to shadow detail being so if it is underexposed.
I'd personally opine that allowing one's attention to be too swayed by fears of shadow noise can potentially work against us, whereas a clipped highlight is, if blown, blown for good. "Noise"(and we'd be having this discussion about grain if this forum existed 20 years ago) has so much potential for doing something with: noise on some sensors can look quite film-like...and indeed, there are relatively easy and manifold ways of multiplying darks so that the "shadow-end" of one's zones soften, fuse or block, leaving the mids and highs retaining an effective high dynamic range.
Yes indeed, "exposing for highlights" certainly...remembering that this does not have to be the same thing as metering from the highlights"... Thank goodness perhaps that we do not all have to use slide(tranny) film and be so very conservative with our exposure.
I confess that sometimes, to "make sure" and cover my laziness, I've metered off well-lit grass, set this as my 18% grey point, then added a stop of overexposure. I'll then make a mental note of aperture and speed at say, f5.6, dial off the correction, nip into my default aperture-priority, blatting away whilst comparing my shutter/aperture with the one I made a mental note. I'll that way(though it's a bit rough and ready) get a mental alarm-bell going if there's a danger of underexposure...sounds more long-winded than it is, but just ensures that I err on the side of overexposure.