You're just rephrasing the "throwaway code" polite fiction. Increasingly, publishable-quality research is expected to be publically reproducible, and that means the code must stay around, potentially in the "long term". Every scientist loves it when their research gets cited a lot, right? Well, those citations become worthless if you can't reproduce the research because the code is an unsurveyable mess relying on bitrotted, unsupported external components.
