01 Check what the rollback changes
A binary rollback can leave the failing configuration in place.
Practice interviews
02 Use current customer evidence
“This delayed detection by nearly an hour.” (opens the source)
A planned change does not explain away an alert.
03 Make the release condition explicit
“Set the evaluation window, failure criteria, and maximum failure count.” (opens the source)
Set a failure condition and an observation window before the change.
“Protect customers from repeated SLO misses” (opens the source)
Apply the service's written policy to its own error budget.
04 Separate a suggestion from authority
“consistent controls to prevent unwanted mutations of production state.” (opens the source)
Check the cited signal and permitted action before changing service state.
05 Keep partial recovery visible
A healthy API is one signal. Queue lag and unmeasured regions still need checks.
Sources
- Matthew Prince, Author, Cloudflare incident report, at the time. Cloudflare Blog, “Cloudflare outage on November 18, 2025”, Nov 18, 2025. Open the source. Checked against the published text.
- CircleCI, Incident report, at the time. CircleCI Discuss, “Post Incident Report: June 23, 2026 - Delays starting Docker jobs”, Jun 27, 2026. Open the source. Checked against the published text.
- CircleCI, Product changelog, at the time. CircleCI Changelog, “Smart Deployments: continuous release validation with automatic rollback”, Sep 2, 2026. Open the source. Checked against the published text.
- Steven Thurgood, Author, example error-budget policy, at the time. Google SRE Workbook, “Example Error Budget Policy”, Feb 19, 2018. Open the source. Checked against the published text.
- Stevan Malesevic and Christopher Heiser, Distinguished Software Engineer and Distinguished Site Reliability Engineer, at the time. Google Cloud Blog, “AI in SRE: Where and how Google is deploying agentic AI to improve operations”, May 28, 2026. Open the source. Checked against the published text.
- Daniel Chatfield and Andrew Lawson, Distinguished Engineer and Senior Staff Engineer, at the time. Monzo Technology, “Tolerating full cloud outages with Monzo Stand-in”, Feb 13, 2025. Open the source. Checked against the published text.
Last reviewed Nov 10, 2026