Service connections can take an unexpected amount of time to close

[clone1018]

Under certain conditions, when a client is reconnecting, there can be a race condition where the client is reconnecting but the services have not finished closing out their connections yet. This causes the stream to still exist in the stream manager state, even though the client is completely disconnected.

I don't think asyncing the service connection closers is the right approach here, because that could cause issues upstream. Likely there's a bug with one of the relay connections and the close call is hanging. If the relay connection hangs, I suspect we can ignore it since it's unique to the particular stream. Need to research more about writing to broken sockets.

I suspect this is related to #20