Did anybody implement in MediaWiki a link rot checker for external links? Here is a theory for how it could work:

When a new page is saved and the parser detects a "http:" pattern, the external link is stored in a separate database table, with a pointer to the wiki page where it was harvested. At regular intervals, all external links are tried (HTTP GET) by a background process and the success or failure rate is recorded. If a link becomes unavailabe (HTTP ERROR) during three consequtive fetch attempts, it gets listed on a special page of possibly broken external links. Broken links from the same website could be grouped together. Maybe the whole site is broken, has moved or has been internally reorganized.