Since there is already a framework for returning letters with invalid/unreadable/damaged addresses, I'd guess the overhead of returning wrongly-written postzegelcodes is negligible.