I would much prefer for the filter to simply send the account/item to be moderated if something bad is detected instead of just blocking it. This allows making the filter more sensitive without it catching half of the english dictionary for supposedly bad words.
If they do make a username filter, it would be kinda hard to discern between “ilikegrapes” and “omgrapes”, and even then we don’t know if “omgrapes” is “omggrapes” but with an elision.