> If you obtain the inode number by fetching the metadata you're already accessing the inode, defeating a part of the optimization Actually this is what fclone

> If you obtain the inode number by fetching the metadata you're already accessing the inode, defeating a part of the optimization Actually this is what fclones does, because it needs other metadata of the file as well (e.g. its length). I've found fetching metadata is lot faster than reading anything from a file, so most of the gains of this optimization are still there.