Outliers are treated fundamentally differently between them, they are treated as bugs in economics, but as features in medicine.

I don't understand what you mean by this.

Let's take for example a simple example of the outlier of the person who smokes a lot of cigarettes but outlives the person who doesn't smoke. Does this break the model where smoking harms health and increases all cause mortality (which we know through epidemiological observation of deaths, which is not in any sense a double blind test)? Where does this observation fit into medicine?

Or take the example of a discontinuity regression in economics. A jurisdiction passes a law increasing the minimum wage above the market-clearing wage in that area, which shares a border with another jurisdiction that has a similar market clearing wage. Can we observe the differences on both sides of that border to see whether the minimum wage increase leads to an increase in unemployment? Yes, it's just applied math at that point.

Where does behavioral economics fit into your ideas of how economics expects a rational actor? There are differences in behavior that have been measured by economists in different situations, and those are important ideas in economic behavior and observations. So why do you assume those models have been discarded in favor of some sort of doctrinal insistence that humans behave in a particular way?

And if you're describing the reluctance of practitioners to abandon the core ideas of their models, or the core paradigms of their disciplines, I'd observe that you're largely correct but wrong to assume it doesn't happen in things that you'd probably call science, from medicine to meteorology to epidemiology. Things get overturned slowly, and sometimes these paradigm shifts meet a lot of resistance for an entire generation: phlogiston proponents slowly coming around on oxygen, cosmologists saying "fine I guess dark energy exists."

The critiques you lob at economics are valid. I just think you under appreciate how much they apply to hard science, too.