The problem is the assumptions made by climate science about the nature of thermalisation. 115 years’ ago, J Willard Gibbs introduced the Principle of Indistinguishability: molecules in an assembly have no memory. Those who teach that trapped photon energy decays over 1000 collisions, the average time an isolated molecule takes to re-emit the photon, teach false physics. This cannot happen because of quantum exclusion..
Because a random thermal emission of a photon occurs simultaneously with absorption, there is on average no local thermalisation so long as the energy can be transferred at near the speed of light to sinks; heterogeneous interfaces and space. So, atmospheric GHGs probably act as a pseudo-scattering energy transfer medium with warming mainly at clouds. The rider to this argument is that above the cloud level, thermalisation rapidly decreases so DOWN emissivity falls to zero at TOA.
That means you don;t need to claim imaginary ‘back radiation’ from the lower atmosphere bounces back to make the Earth’s surface emit at the S-B black body level in a vacuum, a concept disproved experimentally for over a Century. But by eliminating the imaginary ~40% energy increase 100[333 - 238.5]/238.5, it predicts very little if any CO2-AGW!
A rider is that the thermalisation in the ‘PET bottle’ experiment is from pseudo-scattered IR absorbed in the walls. Nahle’s recent Mylar balloon experiment, which shows no detectable warming when you reduce the PET thickness from ~40 mil to ~ 3 mil is attractive but I would like to see it done properly before accepting it as definitive proof of indirect thermalisation.
[The difference of the IR spectrum from clouds and clear sky is a very strong case for the above argument.]