I am using google’s pub/sub to move processing of some tasks (represented by messages on a pub/sub topic) to the background. Some tasks are expected to fail periodically due to known transient errors from a internal service that enforces some measure/minute type limit. This measure is not known or correlated to tasks otherwise it would be much more straightforward to manage. These transient errors are retry-able. Google’s documentation seems to indicate pull subscribers have more flow control than push subscribers.