The model (v0.3.1) had poorly defined boundaries and as a result the downward flux through a layer of the atmosphere was affecting the temperature of an adjacent layer.

I noticed the problems when looking at the model stratospheric temperature profile (the upper atmosphere). According to theory, with no ozone the stratospheric temperature shouldn’t increase from the tropopause. Yet in my model it did. So I ran some stability tests:

different number of layers
different top of atmosphere height/pressure
longer runs
shorter and longer time steps

Flaws kept appearing, including instabilities.

Anyway, the model is now fixed (according to the army of Science of Doom testers). v0.3.3 is logged in the notes.

Because the flaw was in the stratosphere, fixing the flaw (luckily) had minimal effect on the TOA flux results previously reported. Here are the new results: