Academic publishing seems to be messed up but I'm not sure this is a good demonstration of that.

Of the three links that you cite to online copies two no longer link to the article. So while your $32 doesn't buy you anything you can't already get elsewhere the availability (at a price) from Nature does at least supply a degree of permanence to the article. That may be more important with less prominent items.

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.