As identified during discussion in ENTMQBR-3514 - 'Address not created if address CR is submitted before broker instantiated', a more general problem exists. The more general problem is that the operator doesn't go looking for existing CRs upon startup to ensure it is aware of existing instance that may already be running that it will need to manage and update.

This should be addressed as it is particularly important for existing production deployments where a fix to the operator for a bug or CVE may require a slight bump to the operator container image, thus forcing an operator restart. At this point the operator would no longer be in sync with the existing deployment, but rather the deployment would need to be redeployed.

In theory a broker cluster redeployment should result in the exact same deployment with minimal downtime, however steady state would be lost and unnecessary risk introduced.