How does Google decide what to mirror on its servers?

This might be semantics but I think it’s a fairly important difference. Google doesn’t really back up the information they store, or make mirrors of it. Generally both of those would be complete functional copies of the sites.

Google keeps a rough, incomplete, cached copy of the static pages they index while crawling the web. So for example, this page would be stored exactly as it was when Google blazed through here, the community feed would be the saved just as it appeared, however many GQs and GAs were given would stay the same, they may store all the images or not, they may store the styling for the page, or not. Really they’re most interested in the words and hierarchy of the page (what’s a heading, what a paragraph, what’s a link, etc).

As far as how they decide what to keep. I would imagine they keep a page until it is crawled again for the majority of the web. They probably have a threshold for sites they keep more detailed records of. If it goes beyond that I would imagine they delete items that would rank low in the same algorithms they use for their search results, since those are unlikely to show up in any of their products along with any outdated links that no longer appear (once it’s noticed they’re gone).

So I don’t know if the google copy of the web would be very useful since it would be missing so many things. They really only care about the content, and even then only the distinctive content.