Second pair of eyes for #2383: manually re-checked the extracted hanzi against the in-game message font glyphs and their message context. 8 entries were misread lookalikes; codepoints are unchanged so this only affects the human-readable labels, not the extracted bytes.
- ⋯ (U+22EF) → … (U+2026) @0xA109 — the font glyph matches the standard horizontal ellipsis
- 蠃 → 赢 @0xA12E — 000F "你贏啦！！" (bombchu bowling prize); 蠃 is a snail/clam, the glyph re-reads as 赢
- 噓 → 嘘 @0xA5DA — simplified form used throughout the simplified text
- 咩 → 哞 @0xA5F1 — message 2006, a cow talking about producing milk; 咩 is a sheep's bleat, 哞 a cow's moo (牛, not 羊, on the right)
- 亳 → 毫 @0xA605 — 201B "一根毫毛" (a single hair); 亳 is a place name, the glyph lacks its extra upper-right stroke
- 汊 → 汉 @0xA621 — 203F "懶漢" / 404E "男子漢" (lazy fellow / manly man); 汊 means a stream branch
- 鰓 → 鳃 @0xA68E — simplified form (fish gill)
- 寃 → 冤 @0xA75B — standard form (寃 is the variant spelling)
Apply the same replacements to CHN_CHARS in tools/extract_text.py (extraction charmap) so the build matches; charmap.chn.txt is used for building, CHN_CHARS for extraction.