Character set changes depending upon privs

Not sure if this is a bug or feature but try this experiment. On a MariaDB-10 server with a properly configured utf8 database, create a new account and give it all available privs. Now create a second account with all privs except super. Now invoke "mysql --no-defaults" then use the no-super account to connect to the utf8 database. Typing "show variables where variable_name like '%char%' will prove the client is using utf8. Now exit then repeat the whole experiment with the super account where you will learn that you are using latin1. Is this strangeness documented anywhere?