Date: August 17, 2026
Duration: 10:28 AM – 10:55 AM PDT
Impact: Users experienced elevated latency and errors throughout the platform.
The incident was caused by a specific database query used by the version history feed on the process version page. That query reads the entire change history for a process version with no limit and sorts it — in this instance, hundreds of thousands of rows in a single call. During the incident window, this query consumed enough database capacity to create contention for other application queries. As a result, unrelated workflows that depended on the same production database also slowed down or timed out.
We will be optimizing this feature, and have updated our database timeouts as a backstop.
10:28 AM — Specific request to process version history made; database slowdown begins, latency starts to climb
10:31 AM — Database connection capacity exhausted; requests begin failing
10:52 AM — Database-level mitigation applied to update timeouts
10:55 AM — Full recovery