Elevated Webhook Delays

Summary

On September 10, 2025, an incident occurred causing significant delays in Directory Sync webhook delivery. Customers experienced these delays from 14:54 UTC until resolution at 17:46 UTC. The issue was triggered when a high volume of outgoing webhooks timed out during delivery, which led to repeated failure attempts and consumed available webhook worker capacity.

What Happened

During the incident, a customer generating a high volume of Directory Sync webhooks experienced an outage on their system. This led to timeout errors when we attempted to deliver webhooks to them. These timeouts created a growing backlog, which eventually caused delayed webhook delivery for all customers.

During this period, 26 customers experienced delays exceeding 10 minutes for at least 10% of their webhook volume. The 90th percentile delay across the system reached 49 minutes.

Contributing Factors

• A high volume of outgoing webhooks were timing out on delivery to a customer system.
• Timeouts lead to slow failures & repeated delivery attempts, consuming the available capacity of the webhook workers.
• Shared pool of worker capacity for directory synchronization webhook delivery led to a cascading effect that impacted all directory synchronization webhooks.

Timeline

Remediation

Immediate Actions

• Investigated and identified the source of delivery failures.
• Worked to clear the job backlog.
• Monitored the customer system recovery and webhook delivery.

Future Prevention & Improvements

• New Webhook Delivery Architecture: We are testing Directory Sync webhooks on a new architecture that will isolate delivery failures, ensuring that issues with one customer do not impact others.
• Load Shedding Techniques: Identified and will implement load shedding techniques to manage high volumes of webhooks more effectively and prevent system overload.
• Improved Alerting: Enhanced alerting mechanisms are being developed to allow for faster detection and response to degraded webhook performance.

Conclusion

We understand that this incident had a noticeable impact on our customers and their clients. As a provider of essential enterprise functionality, we strive to provide the best possible service, and will be feeding the lessons learned from this incident back into our system. This ensures a resilient platform you can rely on now and in the future.