Page 50 of the report says that some (unnamed) resolvers do this, and “we think this [..] is a very good implementation feature to reduce the impact of mistakes”, but that it’s “definitely not universally implemented”.