From a WebDAV or Chwala or filesystem or (...) resource location, an initial synchronization copies off metadata (of files and folders) -- not payload -- this allows the interface to generate lists for folders and files (available for sessions).
The creation of a session against a document would copy off the data to #Manticore, so that operations and snapshots can be created, until such time the document is (automatically? upon the user's request?) saved off to the canonical location of the original file.