This answer meets the RTO and RPO requirements by using synchronous replication within the same region, which ensures that all writes made to the primary instance are replicated to disks in both zones before a transaction is reported as committed1. This minimizes data loss and downtime in case of a zonal or an instance failure, and allows for a quick failover to the standby instance1.
This answer also meets the high availability and disaster recovery requirements by using asynchronous replication across different regions, which ensures that the data changes made to the primary instance are replicated to the read replicas in other regions with minimal delay2. This provides additional redundancy and backup in case of a regional failure, and allows for a manual failover to the read replica in another region2.

PgBouncer maintains a pool for connections for each database and user combination. PgBouncer either creates a new database connection for a client or reuses an existing connection for the same user and database. + PgBouncer is a simple PostgreSQL connection pool that allows for several thousand connections at a time. Using Kubernetes Engine to run a Helm Chart w/ PgBouncer based on the great article from futuretech-industries, we were able to set up an easily deployable system to get the most out of our CloudSQL DBs without breaking the bank. https://medium.com/google-cloud/increasing-cloud-sql-postgresql-max-connections-w-pgbouncer-kubernetes-engine-49b0b2894820#:~:text=That%20is%20where,breaking%20the%20bank.