An issue that gets far too little attention is the review of models. Models seem to be considered secret intellectual property of the creators, and therefore not subject to inspection by others. As you point out, climate models are complicated, consisting of many parts (modules, let us call them) that must interact with each other, and which include feedback loops that involve multiple modules. How do we know that the code of these modules accurately represents the “known” physics relevant to the module, that there is not additional code and constants representing “informed guesses” about relevant physics, chemistry, etc., where less than complete science is unable to provide the needed information, how scientific or arbitrary the interconnections between the modules are, and a host of other questions many would like to inspect the modules to determine?
It is close to criminal that large, complex bodies of code, mostly developed with government funds, can remain hidden and unexamined by multiple independent investigators. If we are to make major societal decisions about dealing with AGW based on these models, they deserve to be analyzed in depth by independent reviewers, so that we have some basis for believing they are sound. The model builders themselves should be calling for this, because otherwise if society takes actions based on what the models say, they will be responsible for any erroneous actions society takes.
In my view, any climate model developed with public funding should be made available to anyone willing to spend the effort to examine it.