When matching @replies against possible reply targets, we do a certain amount of normalization, so e.g. a user named "Noël" and be replied to by using "@noel". Letters that are neither standard ASCII letters nor treated by this normalization were silently stripped. This is fixed now (the auto-completer already did this correctly, thus the mismatch). In addition, you now have the "honor" of being part of our unit tests :)
Changing your first letter to a latin "T" did not help for this particular issue, since you still had just one "regular" letter, which is not enough for a reply. However it has the advantage that a user with a latin script keyboard can now use their "T" to trigger the auto-completer; however, using a cyrillic "Т" does not work for that anymore. In either case, if the autocompleter offers your name, you will be also notified.
I'm a bit torn with regards to doing cyrillic -> latin remapping (in other words, making both "Tим" and "Тим" work). While for that particular case it would be straightforward, how would we remap the cyrillic "Н" or "Р"? The correct transliteration would of course be "N" and "R", but the majority of the users (especially on sites other than this one) would probably type "H" and "P" instead.