> Lots of computer science research produces useful products that make their own money directly, and so the need for reproducibility is less if impact on the economy is a main criterion.

That research isn't being conducted by academics applying for pure research funding from the Marsden fund...