Very cool. A couple of suggestions, though...
If you save the unconverted HTML in the .html files, (but with relative links to other Wiki pages as e.g. "./SomePage.html") they could actually be served statically as pure documentation. So, if the "cache" were visible directly via a webserver, you'd actually see static doc pages. But, when displayed in the Wiki, you'd have the Wiki wrapper content around the page, including edit links, etc., as long as you stripped off the '.html' in the URLs.
In other words, if you move the '_cleanHTML()' call so it occurs on data retrieved from the cache, rather than on data being put in the cache, this should work. (If you want it to go faster, you might also try using html.rfind(".)