Is is a fundamental HTTP/2 protocol issue or implementations issue? Could this be an issue at all, if a server has strict limits of requests per IP address, regardless of number of connections?