We suspect a "message of death" may be to blame, so we've siphoned off the current queue and restarted workers. Higher than normal error rates persist, and we're gathering more info.
Posted Oct 18, 2015 - 20:32 UTC
Update
We have stopped all OS X worker processes while continuing the investigation.
Posted Oct 18, 2015 - 19:02 UTC
Update
Job throughput is effectively zero, and still no good leads on contributing factors. We are escalating with our infrastructure provider.
Posted Oct 18, 2015 - 18:49 UTC
Update
Job restarts are still much higher than usual. The backlog is growing. We're still investigating.
Posted Oct 18, 2015 - 18:18 UTC
Investigating
We are seeing heightened levels of restarts due to network connectivity errors.