We have a firehose that I would like to write monitoring alerts for. When I started investigating the metrics, I see 40 times difference between between IncomingBytes and DeliveryToS3.Bytes, something like 5 MB input average for 5 mins compared to 200 MB output average for the same period. We do have some dynamic partitioning keys added to the data to partiton by event type and timestamp, but I can’t imagine this increasing data output that much. Do you have any thoughts what might be causing this?