Basically, I can guarantee that killing cookies would lead to 0% adoption of HTTP 2.0 forever, due to the same inertia that is holding Python 3 or IPv6 back. Doesn't matter if there's a better mechanism included that does the same job.