This will make it easier to do caching / keep stats. Also, we won't have S3Client's connection pooling if we create multiple S3Client instances.