Strings that exist in French appear in English on a French page. @optilete mentioned on Crowdin a different (related?) issue with that page in Dutch. This issue should probably not be at the top of the priority list as this is basically a list of institutions and the page remains usable as is @kueda
If you translate everything correct in Crowdin there is no issue. If I ignore everything but the source string the translations is perfect. It means I had to translate everything.
Thanks for the screenshot, @dugald. That page is integrated into Crowdin as a full HTML page, and Crowdin doesn’t do as good a job tracking the identity of strings in an HTML page as it does from a list of translatable strings as we use for the rest of the website. Sometimes this means that when we update part of that HTML page, Crowdin removes the approval for the wrong string and associates strings with translations for other strings. This has been driving me a bit crazy so I changed our settings for these files so Crowdin will delete translations when the strings change in these files. This will make it so people approving existing translations won’t unknowingly approve translations that were correct for other strings but got incorrectly assigned to new strings. Unfortunately it means translators will have to retranslate parts of these pages whenever something changes, even if text was already translated, but I think it’s worth it to prevent weird situations like the one you’re describing here. The “deleted” translations should still be available via Crowdin’s Translation Memory feature.