Reproducibility is part of science. So is identifying and fixing errors. But perhaps the most important aspect of science is being able to continue it.
I've worked in science labs where non-software engineers write code. They fall victim to the same problem software engineers fall victim to when they work without version control: the lose it, they overwrite it, they make mistakes and want to go backwards, they end up with 50 copies and can't remember which one was used to compile their postdoc work. And when it comes time to publish, they (may) archive it and never look at it again - despite the fact that good science should necessitate they release the code, if for no other reason than to reproduce results and ensure they are error free.
Version control is a tool. When used properly, it solves many of the above problems, all which sap productivity. In an academic setting, particularly where peer reviewed papers are being released about computationally intensive science, version control almost certainly saves more time than it creates. People just aren't willing to put the initial investment it takes to learn to use the tool.