The answer to this problem that has plagued me for the better part of a month can be fixed with this command from the start-up directory:
rm -f .sqlfserver.ser
that's the magic file that keeps track of whether SQLFire thinks there is a process running in the data directory, whether or not it actually is. If the system reboots or SQLFire otherwise fails to shutdown correctly this file is not removed. And a "sqlf server start" doesn't appear to check anything (like a process) before it generates the message in the title. A "restore.sh" for the database also fails to notice this file before proclaiming that the database was successfully restored. So, keep in mind that you should always do an "ls -la" to find these kind of programmer cookies.