It is a common situation on Windows when agent collects text data in non-Unicode character set and explicitly converts it to UTF-8 before sending to server.
Luckily, most modern UNIX systems use UTF-8 by default and such conversion is not needed in UNIX agent.
However, there are few deviant platforms on which agent seems to receive text information from the system in non-Unicode encoding if it is run using non-UTF-8 locale. This may influence interpretation of system.hostname, system.uname and other text-based items if they contain non-ASCII characters.
Recommendation to set UTF-8 locale for agent will help in such cases.