We are observing frequent, momentary disconnections from the primary database cluster, primarily between 10 AM and 2 PM UTC. This is leading to 503 errors for users attempting to perform write operations, such as submitting forms or placing orders. The logs suggest a possible connection pool exhaustion or a bottleneck in the load balancer configuration.