Your questions concerning the connection process to read replica instance, depends on how you connect to the instance, you will have to connect to a specific read replica with the appropriate methods discussed in the link above. Additionally, Cloud SQL does not provide load balancing between replicas. You can find a detailed list of operations for configuring replication here. I thought it’d be worth while to explain the concept of read replicas as this information can assist you when deciding whether using a read replica or failover replica is appropriate in your use case.
The idea behind read-replica is to provide some sort of scaling-out for your data. This is done by allowing read operations such as SELECTS on the replica. While other write operations such as UPDATE/INSERT/DELETE occur on the master instance. Due to this, read replicas are designed to provide scalability and not high-availability. If high-availability is what you are seeking you will need a failover replica instead. The failover replica will automatically promote itself to be the master and retain the original master IP. You may explore the benefits of high-availability here.
In regards to read replicas, if your master instance experiences downtime, you have the ability to promote the read replica to be the master. However, this is a manual process and will not be done automatically. In the event this occurs, you may find the appropriate steps to do this here.