A public Cloud Run service performs well under light load but slows and errors under high load during a stress test. How should you troubleshoot and address the issue?

The correct answer is to check and increase the max-instances limit. Cloud Run automatically scales instances based on incoming requests, but it respects the configured max-instances limit. If the service is slowing and erroring under high load, it's likely hitting this limit, preventing it from scaling out further to handle the demand. Increasing max-instances allows Cloud Run to provision more instances, improving performance under heavy load. Checking Cloud Monitoring network traffic for a spike is a good diagnostic step, but traffic splitting to a previous revision is for rollbacks or A/B testing, not for addressing performance under load. Increasing min-instances ensures instances are always running, reducing cold start latency, but it doesn't address scaling issues when demand exceeds the current instance count. Disabling Cloud Armor is generally not recommended as it removes a critical security layer and is unlikely to be the cause of performance degradation unless legitimate traffic is being blocked.