I recently tried to use this extension with an Excel 2010 (.xlsx) file. It basically didn't work because the XML files inside the zip archive have all their data on one line (compressed), making it impossible for Mercurial to do a line-by-line analysis.
I'm wondering if ZipDoc can be enhanced to do some sort of "XML tidy" operation (which would format the XML data onto multiple lines with proper indenting) on Open Office format XML files prior to committing into the repository. This would allow deltas to be calculated correctly.
Also, when updating, the file would need to be re-compressed (untidied) to enable it to open properly in Excel.