Currently the only way to get an operation's job ID from a poller is by examining the status object held by the polling method, for example:

backup_poller = backup_client.begin_full_backup(container_uri, sas_token)
job_id = backup_poller.polling_method().resource().id

It would be more ergonomic and discoverable if the poller itself exposed the ID, or even the entire status object. Looks like such a change would require subclassing LROPoller.