I get these messages from OPM pretty frequently, but I don't understand the metric that throws the alarm. There's lots of CPU, NVRAM is flushing nicely, disks aren't busy ... yet something in the node pair is being over-utilized.
You may already be doing this, but that particular alert is measuring the sum of both nodes in the HA pair relationship. It's alerting you to tell you that if one of the nodes were to fail, you may run into a performance problem until you were able to get that issue resolved. Below is a snip out of the OPM user guide.
Identifies situations where nodes in an HA pair are operating above the bounds of the HA pair operational efficiency. It does this by looking at the CPU and RAM usage for the two nodes in the HA pair. If the combined node utilization of the two nodes exceeds 100%, then a controller failover will impact workload latencies.