In development, you may sometimes refresh the schema. This will remove all virtual datasets. It would be nice to have a selection of virtual (remote) datasets that are then readded to the database with other initial testing data, so that we can easily test on those.
Choose a low number of datasets with different configurations.