I wonder how your system deals with write conflicts? The documentation is not clear about that...
Because of no locks, the last write wins according to rsync update semantics based on file timestamp + checksum. In case two files have the same mtime, rsync compares the checksum to decide which one wins.