The presence of __NOCONTENTCONVERT__ changes how the tokenizer handles language variant markup. It shouldn't be stored inline in wikitext, since that gives a chicken-and-egg problem and makes the parsing of other wikitext on the page position-sensitive. It should be stored out of band as a page property, eg using MCR.
Description
Description
Event Timeline
Comment Actions
Probably worth auditing how NOCONTENTCONVERT is currently used, and whether (or not) we can easily migrate it.
Comment Actions
Obviously this isn't practical for a global usage view, but use on individual wikis can be audited via Special:PagesWithProp, e.g. https://en.wikipedia.org/wiki/Special:PagesWithProp/nocontentconvert.