Google Scholar is doing a reasonable job at solving the OA discoverability problem, though at the risk of us becoming dependent on it.

The deposit problem is, as you say, a lot knottier. I personally suspect a linked network of institutional repositories with a twelve month embargo may end up as the end state of the current process, which I don't think is good enough.

Sci-hub provides a sociotechnical means of doing differently, and that itself justifies its existence, I think.