This is something I've stumbled across before and was also interested in the background to, but never actually raised a topic.
Based on this the recommended should probably be to build up your own test suite for list behaviour to protect against changes in the schema-list (since using
block+ is not adhering to the API).
Personally I'd be in favour of lifting the constraint on schema-list to expect a leading
paragraph, and instead support
block+. Consumers could then be free to impose additional constraints as needed.