Degraded performance for web chat assistants, customer facing conversations, and AI workforce

Between 8:30 AM and 12:28 PM CST, our Conversations service, which powers much of the communication layer our AI assistants and messaging features rely on, experienced a performance degradation that affected message delivery and response times.
Timeline

8:30 AM — Our engineering team detected a minor performance degradation in the Conversations service and began investigating.
9:30 AM — As the impact grew, we escalated the issue to a full incident response.
Root cause — We traced the degradation to session exhaustion in our primary database for Conversations.
Response — We ran two workstreams in parallel: scaling our database capacity through additional replication, and increasing the session throughput available to the Conversations service. At the same time, we tuned our workflow orchestration layer to remain highly available and maintain throughput during session-constrained conditions.
12:28 PM — Service was restored, and the availability alert cleared.

We apologize for the disruption. A full post-incident review is underway, and we'll be implementing changes to prevent a recurrence.