Arvados

This is still an issue, but it may be lower priority: we just realized that #6048 was happening because websockets simply weren't running on the clusters in question.

In addition, we could provide a fallback method that uses log table polling to get events. The Python SDK uses this to great success when websockets isn't available.