Thirteen minutes. A Shutterstock service went down at 12:35 p.m. and came back at 12:48.
The immediate report said a configuration was missing during the rollout of a new deployment system. Development and QA testing had not caught the problem. The follow-up work would include making the development, QA, and production environments more consistent.
My reply asked for a clearer explanation of why the issue had not been caught before production.
That question was specific to the deployment path. If a change works in development and QA but fails in production, one possible cause is that the environments do not have the same configuration. A test can pass while still missing the condition that customers will see.
The February planning emails had already named uniform environments as an important technology goal. The May incident gave the team a concrete example of the cost of those differences. A missing configuration in the production path caused a customer-facing interruption even though the change had been tested elsewhere.
The archive source does not establish how many customers were affected, what the missing setting controlled, or the permanent fix. Those details are intentionally omitted. It also does not support blaming an individual engineer. The relevant record is the system gap between environments and the question asked after the service returned.
Shutterstock reported strong first-quarter results two days later. Revenue had reached $51.1 million for the quarter, the collection contained 25.1 million images, and customers had completed 22.3 million paid downloads. The public numbers showed the scale. The thirteen-minute outage showed one part of the work required to support it.
This is not a complete incident report. It is a dated skeleton built from the initial operational email. A fuller article would need the deployment change, monitoring timeline, customer impact, remediation ticket, and confirmation that environment parity improved afterward.
For May 6, the supported facts are enough: thirteen minutes of downtime, a missing configuration in a new deployment system, a test path that did not catch it, and a request to explain why.