SQL Server CPU Saturation - Claim Submission Failures

Summary

On June 23, 2026, DentalXChange experienced a service disruption affecting claim submission, ClaimConnect, eligibility services, and related platform functions. The root cause was a reporting data-load process that ran uncontrolled on the database tier supporting these services. The process drove SQL Server CPU utilization to 100%, creating cascading database locks that blocked normal transaction processing across the platform.

We understand the impact this disruption had on your operations and are committed to transparency in our response. The corrective actions described in this document address both the immediate performance deficiency and the systemic monitoring gaps that allowed the condition to persist undetected.

What happened

Post-incident analysis by our database team identified the technical root cause: the process repeatedly performed a full scan of a very large audit table (hundreds of millions of rows) that lacked an index supporting the query, making it far more resource-intensive than all other concurrent activity combined. A secondary code-quality finding in a related procedure was also identified and is being remediated.

The disruption lasted 58 minutes (7:05 AM to 8:03 AM PST). Resolution required stopping the runaway process, refreshing SQL Server statistics, and restarting the affected servers to restore service. Multiple external clients as well as internal DentalXChange teams were impacted. This was a service-availability event; our investigation found no indication that any protected health information (PHI) was accessed, exposed, or improperly disclosed.

Root cause

A reporting data-load process executed on the database tier and ran uncontrolled for approximately one hour without triggering an automated alert. The process calls a chain of child routines, the most impactful of which repeatedly performed a full scan of a very large audit table (hundreds of millions of rows) because no index supported the query. This made it far more resource-intensive than all other concurrent activity combined. The sustained load saturated SQL Server CPU at 100%, generating cascading database locks and blocking normal transaction processing across the platform.

A secondary finding identified a related procedure whose join logic was not fully constrained, which increased its resource cost. This is a code-quality issue; a code review is underway to correct the logic.

How we restored service

1. Runaway reporting process identified on the affected servers; blocking queries stopped.
2. SQL Affected servers restarted to clear residual lock state and restore service.
3. All systems confirmed healthy at 8:03 AM PST. Public status page updated. Senior leadership notified.
4. Post-Incident Review conducted. Tracking tickets were created for index optimization, code review, and monitoring enhancements with assigned owners and target dates.
5. Server statistics manually refreshed to support query plan optimization.

What we are doing to prevent recurrence

• Add a nonclustered index to the large audit table to support this query pattern. Validate in the pre-live environment before production deployment. This single change is expected to eliminate the dominant table scan and dramatically reduce runtime. Target completion: 8/7/2026
• Add supporting indexes to two additional high-traffic tables referenced across multiple reporting routines, as part of the same indexing pass. Target completion: 8/7/2026
• Review the related procedure to correct the join logic identified above. Determine the correct logic, remediate, and test in pre-live. Target completion: 8/14/2026
• Configure a new SQL Server alert to flag queries running longer than three minutes, routed to our alerting platform for direct notification to the IT team, removing dependence on user-submitted tickets to detect runaway processes. Target completion: 7/10/2026

Questions

For questions about this incident or to request a technical briefing for your team, please contact your DentalXChange account representative or reach our support team through the usual channels.