IMO it's generally good advice to be wary of rolling too many disjoint use cases into one abstraction, and I think that's the rationale behind recommending poly

IMO it's generally good advice to be wary of rolling too many disjoint use cases into one abstraction, and I think that's the rationale behind recommending polymorphism, but the ergonomics of actually writing OO code make it impractical advice to try to follow as a rule. Inversion-of-control is often a more practical way to solve the same problem, in my experience