This is simply a consequence of the fiscalization of academic values, and doesn't just apply to libraries. Professors need "high-impact" papers to justify their grants so they can get more grants. The fiscal imperative to acquire grants is very strong, because grant overhead is a major revenue stream for the host institution (probably THE major stream.) It's much easier to have a high-impact paper if you publish in a famous journal, so everyone shoots for Science, then Nature, and on down the hierarchy as their field sees it. And they will eat just about any kind of shit to get published there, including having their papers locked behind a paywall. Because they are plugged into a system where getting grant money takes priority over advancing knowledge. Don't get me started on how this skews research priorities and experimental designs...