In this case, it's different than lift. We can run CFD simulations of wings and get accurate estimations of lift/drag etc. We cannot do that for Hall thrusters

In this case, it's different than lift. We can run CFD simulations of wings and get accurate estimations of lift/drag etc. We cannot do that for Hall thrusters without already having run the thruster, as we need to calibrate the anomalous transport parameters to make the simulation match experiment. This makes the simulations much less useful in design than they are in other fields, which is a real drawback. Another point to make is that testing these devices is expensive. They are low thrust propulsion systems that are designed for thousands of hours of continuous operation. They need to be run in very high vacuum, and for higher power devices there are only a handful of vacuum chambers in the world sufficient for the task. Additionally, Xenon is expensive (dramatically moreso since the Russian invasion of Ukraine), so developing better predictive simulations is critical.