A video transcoding platform sees overnight queue backlogs and throttled API calls that on-call discovers late. Leaders want early warnings, quota visibility, and automated remediation without manual console fixes. Which health monitoring, capacity planning, and operational response design best applies CloudWatch, EventBridge, Systems Manager, and quotas?

The correct design succeeds because it connects proactive detection to automated remediation for all stated needs: CloudWatch alarms on queue depth give early warning of overnight backlogs, quota metrics give visibility into throttled API calls, and EventBridge routing alarm state to Systems Manager Automation scales workers and requests quota increases without manual console fixes, uniting monitoring, capacity planning, and response. The SNS notification approach with manual approval for scaling fails the automation concept because keeping a human approver is prudent for controlled changes but leaves overnight queues waiting for on-call action and violates the without-manual requirement. The queue-only EventBridge automation approach fails quota management because adding workers and restarting transcoders handles compute health yet ignores Service Quotas monitoring, so throttling persists despite healthy capacity. The scheduled nightly automation approach fails event-driven detection because running capacity checks on a timetable is valid for predictable peaks but cannot provide early warnings for unexpected backlog spikes or mid-cycle throttle events. As an exam caveat, focus on discriminator phrases like without manual fixes and throttled calls to separate fully automated alarm-driven designs from manual-approval, partial-metric, or schedule-only designs. As a practical check, confirm CloudWatch alarms for queue and quota, EventBridge rule invocations, Systems Manager Automation execution history, and Service Quotas request status.