At nowiki there is a filter for AbuseFilter that tries to detect missing sources. During translation this filter will report fragments as "without sources" even if the article as such has sources. The root cause is that the filter assumes it process a complete article, while it only holds a small fragment of an article. This is quite annoying.
What we need is to detect in a reliable way (or been informed) that "this is a fragment and not a complete article". The best detector so far is to check if there is a newline separating two non-spaces, but this fails on some (weird and probably unwanted) articles.
Perhaps it could be implemented a flag in AbuseFilter that says "fragment" when the text isn't a complete article, and "article" when the text processed is a complete article?
This is also partly the problem observed at T134211