We need teams of people working on decentralizing any important information sources like this, with mirrors and open protocols that don’t have single points of failure
When I had access to academic journals through my (western, wealthy) institution I would still mostly use scihub.
The simple reason was that I got conditioned to doing that because the login wall to doing it legitimately was often too painful. At a certain point my workflow switched to checking scihub first, then going the legal path if scihub failed.
It saves a lot of time, and I suspect it is a lot more common than most people think.
Acquiring a massive library of unlicensed IP, the best-known such library in the world, would also seem to destroy arXiv's credibility with those in power and I expect would be the end of arXiv.
[0] https://annas-archive.org/