Not immune, but pretty reistant, yes. The results of these models are usually compared with actual data from the real world, very much like in astronomy or geo

Not immune, but pretty reistant, yes. The results of these models are usually compared with actual data from the real world, very much like in astronomy or geology. If a model doesn't agree with reality, it gets discarded pretty quickly. In fact, that is what the researchers did in this case: They applied several statistical models of weather extremes to both models and real data and compared them. Every model that failed to predict the actual measured heat wave went out. That seems like a good way of establishing a baseline, and gives high confidence in the predictive power of the remaining models. You could argue about how the weather data was selected, and how it was compared with the models. But that is already a discussion about details of the specific study. This one is a collaboration between multiple reputable institutions, so it is pretty safe to assume sound practices. Notably, the fact that you actually know about statistical abuse in science shows that the scientific process is working as intended: Results are public and therefore can be and are scrutinized. More important stuff also receives more scrutiny, less important stuff can go unchecked until it is cited more often. But eventually errors will be unearthed and accounted for.