Several servers spiked in workload at the same time causing CPU resource contention. We've spread out the load across the other nodes which stabilized performance.
