You cannot experimentally verify oceanic heat absorption, the same way you cannot experimentally verify anything predicted by climate science - because the scale of such experiments make them effectively impossible.
At best you can experimentally verify bits of theory on small scales - say, testing the temperature of a greenhouse with varying levels of CO2 - then combine those bits into a predictive mathematical/computational model, make predictions, and then perform measurements on the planet and look for results within a confidence interval.
But you're taking a problem with a massive number of degrees of freedom and attempting to reduce the search space to a few dozen. This reduction is non-unique, which is to say there are a large number of ways to model a complex phenomenon like climate which will agree with past and present measurements, but fail to account for the actual drivers which will influence the future.
This is partly why all climate models to date have been incorrect, and the fact that they consistently overpredict temperatures is a strong indication of institutional bias.
>How else do you adjust models if not by finding processes previously not modeled, verify them and then model them?
Excluding potential bias from researchers, there's nothing particularly wrong with this approach, it's the best we can do and given enough time and resources eventually such a search should theoretically stumble upon a correct, complete model, though this isn't necessarily practically achievable for many real world problems. But the issue here is that the confidence of such results should not be misrepresented, and that's the part that people who claim the "science is settled" are missing.