Explain a regression when the image digest did not change.

Instruction: Build a release comparison that includes the state capable of changing behavior.

Context:

A fictional API has the same image digest before and after an error spike. A configuration rollout changed a timeout from 800 ms to 80 ms, and a feature flag enabled a new downstream call for 25% of users. The database schema is unchanged. The flag system has an audit log but updates propagate gradually.

Updated

Official answer available

Read the opening below, then unlock the full answer and practical guidance.

The digest establishes the binary identity, not the complete behavior. I would correlate configuration and flag revisions with affected requests, separating flag-on and flag-off cohorts...

Upgrade to view official answer

Related Questions