The w_2017_38 weekly-release is still in progress. In appears that there may have been some sort of network lurch over the weekend that interrupted the ssh connections between the master and the osx build nodes. Jenkins reconnected via ssh and was displaying the tarball build pipeline script fragment as running on an osx node, but not progress was being made. The tarball builds needed to be manually aborted so the weekly-release pipeline would retry.
I suspect that setting timeouts on the tarball builds would avoid this. Also, splitting the tarball builds up into one build per python major version/os/conda env, would make retrying vastly more efficient.