A SaaS company deployed an Amazon API Gateway REST API with Lambda integration across multiple Regions in the same production account. Customers pay for a tiered capacity measured in API calls per second; the premium tier allows up to 3,000 RPS and customers are identified by API keys. Several premium customers across Regions report 429 Too Many Requests on various API methods during peak hours. Logs show the Lambda function is never invoked. What is the likely cause of the 429 errors for these customers?

The 429 Too Many Requests errors, occurring before Lambda invocation, indicate a throttling mechanism upstream. Since the issue affects premium customers across regions and is tied to API calls per second, the most likely cause is that the company has hit its Amazon API Gateway account-level throttle limit. This limit applies across all APIs in a given AWS account and Region. Even if individual customer usage plans allow for 3,000 RPS, the overall account limit can still be breached, leading to 429 errors. Lambda concurrency limits are incorrect because the logs show the Lambda function is never invoked, meaning the throttling occurs before the request reaches Lambda. The API Gateway default per-method limit is incorrect because the problem is described as affecting "various API methods" and is tied to overall customer capacity, suggesting a broader account-level issue rather than a specific method's default limit.