OpenShift started to use up to 4 threads in northd [0] about a year ago. This BZ is to add the capability to RHOSP to set the number of threads and research a meaningful default and how increasing this number affects performance and memory/CPU utilization. [0] https://github.com/openshift/cluster-network-operator/pull/1471
Please make sure the corresponding Jira ticket for OSP 18 is reported, so that we don't lose the tweaks / knowledge gathered here there when triple is gone. Thanks.