Race condition when shutting down

When running either in the development cluster or MN, sending a shutdown RPC to cargo seems to hang the master process.