Issue #7 discussion uses arguments based on a schema.org property domain, but schema.org itself does not define property domains in the formal sense of rdfs:domain. Actually schema.org loosely defines "expected" or "suggested" or "allowed" properties for a given type. At http://schema.org/docs/gs.html#schemaorg_types one can read that a type "has" properties.
The semantics of rdfs:domain is harder : bearing a property entails to be an instance of the class defined by the property domain.
The interpretation of the schema.org type-has-property declarations in terms of rdfs:domain is an assumption made by rdfs.schema.org, but do we agree this is correct?
Just stumbled upon this issue via Google, so don't know whether it is current... but Schema.org's official OWL version does use rdfs:domain to make domain constraints explicit.