Particularly, in the namespacing.
https://www.mediawiki.org/wiki/Help:Images and parserTests list some variations, such as,
File:Example.jpg|''italic caption'' Example.jpg|on page "{{PAGENAME}}" Image:Astronotus_ocellatus.jpg|''[[commons:Astronotus ocellatus|Astronotus ocellatus]]'' (Oscar)
How common are these forms? What sort of dirty diff'ing are we looking at if we roundtrip these in a normalized form?