A more useful thing to do than update the documentation, might be
actually updating the definition.

Not "per definition"; the changes are not in the definition. The
phabricator ticket gets registered as a def update, thence patched and
logged.

Discussed with Oliver: on each wiki listed in the https://meta.wikimedia.org/wiki/Wikipedias_in_multiple_writing_systems#With_Automatic_Conversion_System page except for the Anglo-Saxon one, there is a drop-down on the third tab at the top of the page, allowing for changing language.
Using that feature, here is what I get as patterns of URLs :

• Chinese Wikipedia: /(wiki)|(zh-(cn|hk|mo|sg|tw))/
• Serbian Wikipedia: /(wiki)|(sr-(ec|el))/
• Kazakh Wikipedia: /(wiki/)|(w/index.php)
• Kurdish Wikipedia: /(wiki/)|(w/index.php)
• Inuktitut Wikipedia: /(wiki/)|(w/index.php)
• Anglo-saxon Wikipedia (only runic language): /wiki/

Given the current regular expression for matching url first folder, it seems we count at least enough pageviews, and might count even more pv than expected for china.

"^(/sr(-(ec|el))?|/w(iki)?|/zh(-(cn|hans|hant|hk|mo|my|sg|tw))?)/"

China

Why are we overcounting Chinese?

I'd say so:
we match zh(-(cn|hans|hant|hk|mo|my|sg|tw))? while, with the language drop-down we saw yesterday, I can only access zh(-(cn|hk|mo|sg|tw))?

Well, https://zh.wikipedia.org/zh-hans/ and https://zh.wikipedia.org/zh-hant/ - they exist, they're just presumably not common enough to be explicitly linked that prominently(?) we should poke a zh-wiki user.

Discussed in standup this morning with the team : we are gonna live it as it is now.
zh-(hans|hant) folder were not present in pageviews on the 2015-102[16-22] so it's not worthwhile changing the definition.
Documentation has been added to the pageview definition discussion page.