In the VSCode codebase, this list is more complete.

wordWrapBreakAfterCharacters: register(new EditorStringOption(
EditorOption.wordWrapBreakAfterCharacters, 'wordWrapBreakAfterCharacters',
// allow-any-unicode-next-line
' \t})]?|/&.,;¢°′″‰℃、。｡､￠，．：；？！％・･ゝゞヽヾーァィゥェォッャュョヮヵヶぁぃぅぇぉっゃゅょゎゕゖㇰㇱㇲㇳㇴㇵㇶㇷㇸㇹㇺㇻㇼㇽㇾㇿ々〻ｧｨｩｪｫｬｭｮｯｰ”〉》」』】〕）］｝｣',
)),
wordWrapBreakBeforeCharacters: register(new EditorStringOption(
EditorOption.wordWrapBreakBeforeCharacters, 'wordWrapBreakBeforeCharacters',
// allow-any-unicode-next-line
'([{‘“〈《「『【〔（［｛｢£¥＄￡￥+＋'
)),

Thank you, that list is longer than that I expected, and some characters are to wide to use the overflow strategy. I need to come up with a more robust method rather than editing on current one. Stay tuned.

Demo: Added a text input to word wrapping section to test word wrapping without recompiling the code

Current code lacks optimization and seems to report false real line widths (See Demo -> Text -> Word Wrapping, the boundary takes space into consideration)

and buggy...

I found why my behavior is different, my algorithm wraps at the space before words, the old algorithm wraps at the end of words. The rendering result is same.

Wrap candidates before:

This| is| a| sentence.|

Wrap candidates after:

This |is |a |sentence.|

Also fixed issue #8503 and #8139, implemented pull #8439