Opensearch ingestion pipeline is idling

New Opensearch ingestion pipeline starts idling for 30 minutes after 5 minutes of processing time and repeats this pattern. We are using SQS and S3 as sources, and there are no error logs in CloudWatch. Currently, there are >4000 messages available in SQS, but the pipeline is not processing them (0 messages in flight and no CPU/memory activity in the pipeline). We observed this behavior today but not until Monday (02-12) as we had multiple successful pipelines using the same configuration with different index names. What can be the issue? Is this a degradation on the AWS side?