Problem: In category pages, pages with titles that begin with a kanji are currently sorted in Unicode order and grouped by the first character in the title. This is not a traditional way to sort and group things in CJK languages, and it’s not very helpful because Unicode order is opaque and does not correspond to anything in any CJK language.
Proposed solution: Titles that begin with a kanji should be grouped the traditional way, by stroke count (e.g., the kTotalStrokes field in Unihan_IRGSources.txt if using Unicode 14), then within any stroke count grouping, sorted by Kangxi radical order (e.g., the integer part of the kRSKangXi field in Unihan_RadicalStrokeCounts.txt), any ambiguities that remain can be kept in Unicode order.
Who would benefit: native CJK readers who use category pages
More comments: I read a proposal to sort these by pronunciation, but that would not work because every kanji is pronounced differently in different languages (and there can be more than one pronunciation even within a single language), so any pronunciation chosen would privilege one language over all others.