They can’t get grants, write papers, or get tenure by mucking about with source code control systems, etc. They need verifiable results, as I assume they’re trying to do things on the up and up. Anything that takes away from thinking things up, publishing them or talking about them is overhead to be avoided wherever possible. Their 2nd year grad students might think of these things, but they’re busier than the senior researchers, what with coursework and teaching assistantships. And they’re very allergic to paying for anything related to software.

What is ideal behaviour for commercial software is almost incomprehensible to these folks. Spend time writing tests?! Fix an input process that’s very long? Pah, grad students have lots of time to do it.