same boat, smaller scale. been hitting overloaded errors sporadically for the past week. switched one of my pipelines to the AWS Bedrock endpoint and it's been solid. not a permanent fix but good enough to keep moving.