One file (4th Injury-Paper.doc) has been found to produce a different output each time it is normalised or converted with Xena. In this case it is caused by the underlying conversion (as done by LibreOffice version 3.3.3). Xena should produce the same output when run on the same file wherever possible. In this file one of the elements in the footer is the full name of the file, which could be one cause for differing results (although in this case LibreOffice is aware that this is a file name). Note that this particular file produces different output when doing multiple conversions immediately after one another using the same input file.
This also brings up the issue of other situations where it may be possible that re-normalising will produce a different result. This would seem particularly likely for documents with content that is not fully supported for conversion (which might result in something like converting formulas to static values) particularly if that content is metadata or dynamic content that is likely to change (e.g. today's date).