So with irrational optimism I went back to google and found the same crappy stuff. (Why does everyone want to redo the same contrived machinations with DOM?) But in a dojo post I found a reference to JSXML, a cross-browser library that includes REXML, using regular expressions to parse a document. While it hadn't been updated ina while it touted speed as a feature and the code was fairly small.
Plugging it in was a piece of cake and amounted to building my useful object from the tree of parsed objects instead of as-you-go. It may be less efficient, but it's much easier on me.
 While the REXML boasts about being super-speedy, it also hasn't been updated in four years. Browsers may have made strides in XML parsing since then.