How odd, must be the default platform encoding. Which version is this? can you switch to debug and attach some comparable log messages to a bug in jira.jboss.com so we can take a look (both windows and linux). Also type "set" at the command prompt on linux and output that to a file and upload it (I'm specifically looking for things relevant to character sets and NLS).
I'll see what I can do... I'm currently working abroad so I have only little time each day to test things. But I will try ASAP to give you feedback.