Your application demands frequent connection and disconnection from the database. You have three listener processes that are listening for the database PROD. While setting up the connect string…

Using only the first address sends every connection to a single listener, creating a bottleneck and completely negating load balancing across the other listener processes.

Selecting one address at random (LOAD_BALANCE=ON without failover) distributes each new connection attempt to a randomly chosen listener, spreading the load evenly across all available listener processes.

Trying addresses in sequential order always attempts the first address first, resulting in uneven load concentration on the primary listener rather than distributing connections across all three.

Trying each address randomly until one succeeds (LOAD_BALANCE=ON with FAILOVER=ON) combines load balancing with resilience - it randomly picks a listener and falls over to another random choice if needed, ensuring both distribution and connection reliability.