Confirm user impact and assign an owner
Check the service from outside your own network and identify which journeys fail. Assign an incident lead, record the start time and pause unrelated changes. Compare the incident with recent releases and provider status information before committing to a diagnosis.
Keep monitoring and incident notes in a place that remains available if the application goes offline. Share a short, factual status update with the affected team.
Escalate with evidence
Contact your hosting, network or protection provider through the agreed incident channel. Share the affected addresses or hostnames, timestamps, symptoms and relevant traffic measurements. Use the provider’s secure channel for sensitive logs.
Ask which part of the path is constrained and which mitigation can be applied there. Repeatedly restarting an origin will not resolve congestion outside that origin.
Change one control and observe
Prefer established protective policies and provider guidance over broad emergency blocks without a rollback plan. Record each change and compare successful user traffic before and after it. Check APIs and monitoring probes as well as the homepage.
If a control causes widespread false positives, adjust or roll it back while coordinating the next step. The immediate objective is restoring useful service, not maximizing the number of blocked requests.
Verify recovery and follow up
Confirm recovery over a sustained observation window and across the user journeys that failed. Preserve logs according to your retention policy, capture the timeline and document remaining uncertainty.
After the incident, review detection time, escalation delays and policies that helped or harmed availability. Assign concrete improvements and test them in a controlled exercise. An incident report should make the next response easier.