Capacity-aware inference: Automatic instance fallback for SageMaker AI endpoints

Amazon SageMaker introduced capacity-aware inference that automatically falls back instances to handle AI endpoint requests efficiently, improving reliability and scalability.