This continues to bother me about start_date:cause and end_date:cause. If I happen to be able to explain how something was created but have no idea when it was created, then there would be a dangling start_date:cause without a corresponding start_date. This situation is underscored by the new EDTF fields in iD: iD strongly associates start_date:* subkeys with start_date in the inspector.
A proposed import of U.S. county boundaries would include a full sentence in the start_date:cause of some 17,000 more features. I think most of these sentences would fit right in with a start_event key. After all, the field in the import source is called CHANGE, so it’s really about an immediate, instantaneous event that “causes” us to create the relation, rather than a phenomenon that indirectly gives rise to the boundary. And either way, it definitely isn’t what caused the date to happen – that’s just the inexorable march of time.