QBO's reconciliation loop is just one worker. This is okay for most deployments, but for OpenShift clusters where there are lots of namespaces (500+), reconciliation can take quite a while to happen due to the number of namespaces and the number of API requests that QBO needs to do for each repository (6 or 7 in total). Further, QBO doesn't seem to refresh pull secrets properly in certain cases where secrets for already created robot accounts already exist. The only way to update those secrets is to delete them and then wait for the operator to recreate them.

Two improvements:
1) introduce parallelism in the reconciliation loop by running multiple workers that would do the reconciliation process. This would improve the speed of reconciliation by a significant margin.
2) make sure that robot account credentials in Quay always match the created push/pull secrets, if they don't always update the secrets inside OpenShift to match credentials in Quay for each robot account.

Please check!