.XML files can get quite large (multiple GB's) and loading them in RAM completely using utf8content: might cause new challenges.
Currently it is much worse than that, Everything will load the entire file into a memory stream for the iFilter, then the iFilter reads chunks as Wide Chars, which is then copied to a single large UTF-8 buffer for the search to execute. This could end up using 4 times the size of the original file in memory.
iFilters are designed more for indexing (removes formatting and styling), whereas Everything is more interested in searching files as raw text.
I would prefer the ability to search xml files for key="value" as your would perform a search in a text editor, so I am investigating an option to override some ifilters (txt,html and xml) for performance reasons, accuracy and consistency.
I'll add an ifilter-content: search to so the user can specify ifilter content searches only.