Should tools/zh_surnames.py take a mainland-weighted third ranking, so the next 付 is caught without a hand addition? #981

#952 added 付, 代, 候, 芦 and 农 (with 蘆, 農) to locales.ZH by hand: the mainland top 200 by people (NCIIC 2008) that neither the Taiwan MOI nor the Wikidata ranking's top 300 reaches. The list's two sources are regional in opposite directions and neither is mainland-weighted, so the next mainland surname past their cut will again need a hand addition.

The candidate weighed in #952, wainshine/Chinese-Names-Corpus (Apache-2.0), ranks distinct names extracted from text by a word splitter, not people, and its README admits bad entries. Its first-character ranking tracked the registry's closely in #952's comment. After #952 the mainland list leaves 28,960 of 1,144,226 names whole (2.53%), spread thin over a tail (芮 588, 容 521, 戎 516, 楚 501, 国 435, 原 421 …).

Open questions: is a text-derived ranking acceptable as a source; where would its cut sit, given that the tail past 300 is where Japanese surname openers sit (山 is 347 misses in that list); and does the tail matter at ~2.5%?

Pointer: decisions/W1's Open line (2026-10-30 #952 entry).