Comment 6 for bug 1906372

The OpenStack team also suggested a similar approach to tenacity.retry, where connection timeouts may be retried once (or more if an arg like "--connection-retry" could be passed).