Emails I have (happy to share if you want) contain XHTML, as one part of a multipart email. Prior to this pull request, the priority on the application/xhtml+xml magic detector was 50, equal to the priority on the message/rfc822 detector. Because of the relative position of the two detectors in tika-mimetypes.xml, the emails were incorrectly detected as XHTML documents.
With this PR, by downgrading the priority of application/xhtml+xml to 40, the more-sensitive email magic detectors take precedence, causing the emails to be properly detected as message/rfc822.
I have not run this thru the govdocs tester or anything other than my own documents, so, full disclosure, this could cause false negative xhtml-detections elsewhere.
I should note this occurs on trunk, from Github, up-to-date as of Tuesday-ish.