Currently, the job connects to the gateway with a reverse SSH tunnel. This approach makes it harder to detect service failure, on gateway failure all connected services become irreversibly unavailable.
This issue suggests establishing an SSH tunnel from the gateway.