When building, testing just one language, not all tests are applied. maybe it is better to do so; had to do a lot of trial and error this morning because of it to get rare characters working.
By the way… I still feel the characters allowed in words depend on the tokenizer and so the language. I think it is better to move the code to the language, or allow all valid language characters, not just a selection.
Please write down the tests that are not applied. It’s a bit of manual work to make them work per-language instead of globally. Plus we need to solve the issue first that the tests then run twice (which makes running all tests much slower).