During a blue/green deployment of our ingest service, load on the underlying Docker cluster increased. The elevated load caused communication to our downstream message bus to slow down resulting in container cluster instability. This lead to rejected HTTP requests and check failures for a subset of elements. Though it was later determined this issue was not related to a code change, we immediately rolled back the deployment as a precaution. We have reduced overall load on the cluster to provide more headroom during future ingest service deployments, and plan to deploy it in smaller batches moving forward.