HttpListener Stop() does not properly clean up.

I am observing the following behavior. First I set up an HttpListener object like this:

            _listener = new HttpListener();
            string prefix = String.Format("http://*:{0}/", port);
            _listener.Prefixes.Add(prefix);
            _listener.AuthenticationSchemeSelector = SelectAuthScheme;
            _listener.Realm = realm;
            _listener.UserCredentialsFinder = GetCredential;

The I start the listener and begin listening asynchronously

            _listener.Start();
            var getContext = Task.Factory.FromAsync<HttpListenerContext>(_listener.BeginGetContext,   _listener.EndGetContext, null);
            getContext.ContinueWith(t => ProcessRequest(t.Result));

I then stop the listener:

            _listener.Stop();

I then start the listener again (either after a set amount of time or immediately, it doesn't matter) and I get back "400 Bad Request" on the next request

The issue is that the HttpListener stop method does not properly clean up. It will stay listening on a stream read here so the next time a connection comes through, the listener won't have the prefix anymore since it was removed in Stop() and BindContext will fail because searchListener will no longer return a valid listener here. I was able to stop this behavior by adding in a cleanup(false) call at the end of Stop(). Is this appropriate behavior? If not, then what can be done?

Start with the HttpConnection.cs stream-read location and EndPointListener.cs BindContext/searchListener location linked in the report, then trace how HttpListener.Stop() removes prefixes and performs cleanup. Reproduce the sequence of Start, BeginGetContext, Stop, and restart, and compare behavior with and without cleanup(false). Done means restarting the listener no longer produces a 400 response from stale connection state.