The illegality (or quazi-illegality in the case of the links from the edu sites) of the mirrors is what's keeping them from being hosted permanently on a free s

The illegality (or quazi-illegality in the case of the links from the edu sites) of the mirrors is what's keeping them from being hosted permanently on a free site. If the copyrights were clear, there would be repositories like arxiv.org that would ensure high-availability free access. The paper isn't on arxiv.org currently because Nature uses its copyright to ensure that it's not.