Currently addprocs() does not handle this kind of machine spec: ("workerhost 10.1.1.1:10000", 2)
The problem is that each worker on workerhost tries to use the same bind port, which fails since only the first worker can have that port. Instead the extra workers should pick an available port starting from the bind port. launch_n_additional_processes() and/or launch_additional() will need to be modified to support this.

Reported on slack.