All the functionality that WWW::Mechanize::Firefox has for (not) using the cache can be found by searching its documentation for the keyword "cache". For example, the ->get method has the optional parameter $bypass_cache. Does that help your first question?

For the loading and/or linked resources, I don't know, but I would look at how the Mozilla Live HTTP Headers do that, and then replicate that logic either in Perl, using WWW::Mechanize::Firefox and/or MozRepl::RemoteObject, or in Javascript. Note that the progressListener interface as documented by Mozilla might help you. It even has methods to access it in WWW::Mechanize::Firefox. Maybe you can help me make the documentation more accessible.