Yes, but to me the important difference lies in what proportion of that learning curve is conceptual/inherent in the domain (knowing what 100 filters and their

Yes, but to me the important difference lies in what proportion of that learning curve is conceptual/inherent in the domain (knowing what 100 filters and their parameters, or 25 different layer blend modes do; Learning what filter you need to apply to get from some initial state to some desired state, etc. In some cases too much of the learning curve is caused by bad design (not discoverable, odd feature grouping that is there for no other reason than the fact that someone cobbled it together that way in V1, ...). Tantacrul's other videos on engraving software UX show some great examples of this, where improvements in UX can help without lowering the productivity plateau for experts (except of course throwing them back if you were to change what they've gotten used to from one version to the next).