This PR fixes an issue where the remote worker was stopped twice and not
properly handled. This adds a synchronised running check to the start
and stop methods preventing closing of a channel more than once.