`kv0/nodes=3/cpu=96` had a serious (66%) regression in terms of ops/s since the Apr 30th run. The regression occurs on both GCE and AWS.   Jira issue: CRDB-27754