We are trying to migrate a on-premise PgSQL (v10) to GCP Cloud SQL's PgSQL (v10) via PubSub replication. I am trying to get familiar with best practices in this context. We built 2 publications - one for all tables with Primary Keys and other for non-Primary Keys. There are few archive tables with non-primary key are having issues in the replication. How do we troubleshoot this? Is it a best practice to bifurcate publications this way?