It looks like migrating to a new data store (from file:// to s3, or s3 to ssh) is as simple as moving the files, and optionally renaming the ~/.cache/duplicity entry to the md5 of the new destination, though omitting the last step will just cause Duplicity to fetch the metadata. Obviously, all the prefix flags will have to match, but otherwise, it looks like migrating is straightforward.
Am I missing anything? My plan is to do an initial backup to a local disk, use AWS Import to move the data to S3, then do incremental backups off of that.