GitHub Actions Caps Large Workflow Run Query Counts to Prevent Search Timeouts

GitHub has implemented a fundamental change in how the GitHub Actions REST API and web UI calculate and report matching workflow runs. When users query workflow runs filtered by parameters such as workflow identity, event trigger, execution status, branch, or actor, result sets that exceed 2,500 records will now return a capped count of '2,500+' rather than attempting an exact aggregation. The platform continues to serve paginated result payloads up to the standard limit of 1,000 items, but it halts exhaustive record calculations past the 2,500 threshold across github.com and GitHub Enterprise Cloud. This update directly targets enterprise teams that operate sprawling automated CI/CD infrastructures running thousands of daily pipelines. Previously, when queries searched through deep historical runs across large repositories or organizations, the backend frequently hit internal execution timeouts. Instead of surfacing an explicit failure, queries often returned the partial tally accumulated prior to the timeout, presenting inaccurate counts to operators. By capping the count at 2,500 records, GitHub trades arbitrary counting precision for query determinism and backend performance. This change aligns with a broader shift in developer tooling and cloud platform design toward resilient API pagination patterns. High-throughput CI/CD systems generate terabytes of ephemeral workflow telemetry daily. As platforms scale, performing unconstrained table scans to return exact counts on high-cardinality metadata introduces severe database load and degrades service availability. Major SaaS providers have similarly transitioned to approximate counts and cursor-based pagination models, establishing strict boundaries between interactive querying and heavy batch data export. In practice, DevOps engineers and site reliability teams must audit existing observability dashboards, compliance scripts, and internal tooling that interact with the GitHub Actions workflow runs endpoint. Any automation relying on precise numeric evaluations from wide, open-ended queries may break or miscalculate velocity metrics. Practitioners should immediately update automated audit scripts to enforce narrower search dimensions—such as applying explicit ISO date ranges or targeting discrete branch names—ensuring match results remain below the 2,500 record ceiling.