The systems provided by publishers are, um, not. They're at the usual quality of bad corporate software, but more strikingly their anti-piracy changes have only served to make the gap worse. There are anti-crawler features which freeze access to entire universities when a hidden link is followed. There are view-in-app and download-disabled restrictions, often associated with no-searchable-text. (Which are especially idiotic, of course, because you can't actually restrict access to static content once you serve it. Using screenshots and OCR is irritating, but not hard to apply.) And in the worst cases, there's "available on request" via a library or some other human-interaction delivery mechanism. Which actually would limit bulk piracy, but only by crippling access altogether.
Every time I've tried to get a background in a topic or write about it, there's been a step where I check a dozen papers and discard 80% of them as being irrelevant for my purpose (or simply bad). Assuming that's common, it's hard to imagine how silos with custom access mechanisms can endure against a single, searchable repository at any price.
It's funny how absolutely useless Google scholar is for finding papers. You can specifically put in authors and the exact title in quotes and it will still be on like the third page.
Point being, if articles can be freely distributed, that paves the way for discovery services to provide better interfaces to the research.