For FileStorage DB, this avoids:
- keeping a lock on the source DB during the whole import,
- saving the whole index when the import was resumed.