I'm trying to gauge if the following is representative of a bigger problem.
Searching for the word "Fjord" on Google with the browser's language preference set to English on google.com yields the zhwiki page as the first Wikipedia result. https://zh.wikipedia.org/zh-cn/en:fjord.
The expected page is https://en.wikipedia.org/wiki/Fjord
Upon querying the search console for this URL, one sees the following:
Page indexing Page is not indexed: Page with redirect User-declared canonical : https://en.wikipedia.org/wiki/Fjord Google-selected canonical: https://zh.wikipedia.org/zh-cn/en:fjord
The page indexing info for https://zh.wikipedia.org/zh-cn/en:fjord says the canonical is https://en.wikipedia.org/wiki/Fjord as expected.
As far as I can tell enwiki-Fjord was never a redirect for desktop UAs. But on mobile, it does 302 to en.m.wikipedia.org/wiki/Fjord and is therefore treated as a redirect. For this reason, the page is no longer treated as a canonical and Google "guesses" at a canonical by looking at page content and concludes that the zhwiki page is the right canconical because it is a redirect to the same enwiki article.
The Google-selected / guessed canonical is apparently based on content match. Understandably the zhwiki URL matches the eventual page that the mobile crawler sees because it itself is again a redirect. The canonical for the zhwiki page is however the enwiki non-mobile version.
I don't know how widespread this is or how problematic it will be. I examined a couple hundred URLs that are marked as "Pages with redirect" in search console and therefore not indexed but could not find another example such as the Fjord one. In other words, all the other pages that Google seems to think are redirects are indeed redirects.
What's special about the Fjord article?
If an attempts to index an en.wikipedia using a mobile crawler are redirects leading to a weird canonical being chosen by Google, why isn't this happening to other pages (even pages that got crawled on the same day with the same crawler)?