Possible performance improvement by avoiding Filtering Nodes by Labels?

I answer a question about efficiency (see link). What occurs to me, is if Neo4J could keep track of which Relationships have unique incoming and/or outgoing Labels for Nodes, then it could avoid Label filter(s). (This would requiring tracking every Node/Relationship creation, deletion, and modification.)