We keep two separate writers, one for the ontology and another one
for regular data. The ontology one is just used during database
initialization and is not expected to survive past
tracker_data_manager_init(). The second writer will remain in place
in order to dump user updates.