A multi-unit managed (2SU) kafka cluster is causing the assumed worst-case bandwidth available to a ingress replica to be much smaller than it should be. For example with just a single 2 unit managedkafka out of a total of 25 units in production, it computed 11 replicas were needed rather than 2.

This is because the current code accounts for the "ingress+egress * the broker replicas", rather than accounting for each broker replica separately.

This will no longer be an issue after MGDSTRM-8908