Hi Thomas, are you running with a large number of fdbserver instances + client instances? Could it be that Cluster Coordinator is bottlenecked on number of monitoring all connected entities?
Please see some recent threads https://forums.foundationdb.org/t/cluster-controller-cpu-utilization-pegged/740 and https://forums.foundationdb.org/t/migrating-from-a-large-cluster-to-another/721 if they give you any clues…