It was a bit of an oversimplification on my part, since for a given wavelength, the atmosphere isn't binary, completely transparent or completely opaque (i.e. t

It was a bit of an oversimplification on my part, since for a given wavelength, the atmosphere isn't binary, completely transparent or completely opaque (i.e. the radiation would go a very very short distance before being absorbed), but for _most_ wavelengths it's _very close_ to binary. Either it's absorbed or it isn't, to a first approximation. Yes, some wavelengths are absorbed very weakly, and some are scattered rather than absorbed. But these represent a tiny fraction of the spectrum, the nuance would've overcomplicated the post, and they don't change the result much. For wavelengths that are not absorbed, they can be reflected back out just as easily as they came in. For wavelengths that're strongly absorbed, they heat the first parts of the atmosphere they encounter, and this is why the thermosphere is called that. For those few wavelengths that're weakly absorbed, yes they produce some heating on the way in, and if you reflect them back out, they do produce some heating on the way out. But as you've said, because the intensity of the radiation is decreasing along the whole path, the "outbound" contribution is weak indeed. It turns out to be basically a rounding-error in the thermal contribution, which is overwhelmingly dominated by blackbody absorption and re-emission, which often does produce wavelengths that're strongly absorbed. On a perfectly-white planet (or snow-covered region), you might have to start caring about weakly-absorbed wavelengths, but for most types of ground cover, other effects swamp it.