Hi Lilit,
I wonder if it's related to the
bug that was submitted last Friday. Other users have been having trouble as well. From my experience, I have usernames in Latin and Arabic script which have both worked fine in the past. Even if there is an error in the usernames, it should validate and then report which usernames are valid and which ones are invalid (allowing you to trouble shoot and check for typos, trailing spaces or unnecessary underscores).
I know our colleagues in the Czech Republic were having trouble with usernames with complex characters with diacritics at first, but it turned out to be a character set adjustment. The quick solution for them was using a Google spreadsheet for the CSV. This is something you can investigate once your cohort validates and you know which usernames the tool accepted or rejected as invalid.
Dan beat me to the bug reply just now :)
Cheers,
Tighe