With both modes, the reading states are stored within the .reading-states directory.
The files and directories are read from a cached portion of the filesystem — the database — built and updated during the import phase, stored in .metadata.json.
The shelf displays the descendants of the current directory.
The files and directories are read directly from the filesystem.
The shelf displays the direct children of the current directory.
You can use plato-import to off-load the import task to a computer.
You can import with plato-import -I LIBRARY_PATH.
If new entries were added, you might populate the metadata with plato-import -a ADDED_DATETIME -E LIBRARY_PATH where the argument passed to -a is the added date-time of the first added entry (the new entries are at the bottom of the database).
You can then edit the database with your text editor to manually fix the metadata.
You can make a backup of a library with:
rsync -vurt --delete EREADER_LIBRARY_PATH/ COMPUTER_LIBRARY_PATH/