How do linear session type frameworks deal with communicating over an unreliable channel (eg the network)? Do they assume that’s handled at a lower level or does every protocol step have to have an explicit disconnect edge to call in that case?