Actually the question is not between fraud and incompetent programming, but whether between scientific fraud and scientific naïveté. Climate modelers should (and some do) understand that because they are dealing with a non-linear dynamical system, long-term prediction is impossible (the popularizing term for the phenomenon is 'chaotic dynamics') and that the coarser the discretization used in the simulation the worse even short-term predictions will be.
It's not a matter of competent or incompetent programming, but of programs, no matter the correctness of the code or the sophistication of the methods used, being unable to predict over the long term.
The "weather isn't climate" refrain suggests that deep down, there is a degree of scientific naïveté underlying the whole thing: they seem to assume that the short-term variability of weather is statistical noise, when, in fact, it is the short-term dynamics of the system they are modeling. If you have a non-linear dynamical system (weather) and make another by taking averages of the variable over fixed-length time intervals (climate) the new dynamical system is still non-linear, still exhibits chaotic dynamics, and is still not feasible to predict long-term.
I have a colleague, who, for a while, was taken in, until he went to a conference on the subject and saw how really bad the mathematical models being used are. Bad models usually are the result neither of fraud nor of outright stupidity, but of not understanding something fundamental about the nature of mathematical models or about the system being modeled.
This is why I've thought the whole thing was a crock from long before the fact that they 'modelers' have been engaging in fraud came to light.
See #37.