Non-ICU unicode support (Wchar) does not have a line breaking implementation beyond simply testing "category(c) != Separator_Line". This causes Asian text to not wrap correctly. Also, it should take into account prior context when breaking lines (see https://bugs.webkit.org/show_bug.cgi?id=105692#c52).
Wchar unicode has bigger problems: it doesn't have basic unicode support, including support for determining the category for a specified category. Without addressing that there's no good way to fix this issue.
I sent a message to webkit-dev a few days ago raising the possibility of embedding limited parts of ICU in order to provide a much lighter weight but functional unicode implementation.
Not much response - just one "I'm interested in this too, any updates?".