When creating a new database (this should in principle be done every new edition of TREC OpenSearch), the proposed workflow is to import a fixture from the dump/ directory:
By importing this database fixture, some users and data are added to the database, also the index hints are imported in the database.
Maybe it is good to come up with a new way to initialize a new database. We could keep the current mechanism, but remove the irrelevant users and data from the fixture. Or, we could abandon the fixture, and come up with a method that just adds the index hints. Index hints for the new 'systems' collections in the database may have to be added as well.