The API servers of the Postmark service experienced a significant outage earlier today, causing connections to drop and messages to be lost. The team quickly investigated the issue, redirecting traffic to an off-site disaster recovery data center to minimize losses. After identifying the root cause as a malfunctioning RabbitMQ cluster, they forcibly killed two nodes to restore the single healthy node, allowing traffic to be redirected back to their Chicago data center. An investigation is underway to determine the cause of message duplication and ensure such issues do not recur in the future. The service's priority is to prevent losing connections or messages, and the team acknowledges that today's outage fell short of this goal.