Split the EventStreams headless service for the Kafka broker and controller into 2 separate services to avoid "duplicate timestamp" issue in OCP kube-state-metrics

After upgrade to Event Streams operator 12 (exact operator version 12.0.2) we have observed that the Endpoint resource event-brk-<env>-kafka-brokers contains duplicate entries for kafka broker and kafka controller, causing PrometheusDuplicateTimestamps alert to fire. 

By the public runbook available for the firing alert we have investigated that the cause for the duplicate metrics is the event-brk-<env>-kafka-brokers endpoint, those are the debug log from the Prometheus pod:

1 ts=2025-10-07T07:33:17.590Z caller=scrape.go:1777 level=debug component="scrape manager" scrape_pool=serviceMonitor/openshift-monitoring/kube-state-metrics/0 target=https://172.17.3.183:8443/metrics msg="Duplicate sample for timestamp" series="kube_endpoint_address{namespace=\"vedoc-core-dev0\",endpoint=\"event-brk-dev0-kafka-brokers\",ip=\"172.17.39.39\",ready=\"true\"}"

1 ts=2025-10-07T07:33:17.590Z caller=scrape.go:1777 level=debug component="scrape manager" scrape_pool=serviceMonitor/openshift-monitoring/kube-state-metrics/0 target=https://172.17.3.183:8443/metrics msg="Duplicate sample for timestamp" series="kube_endpoint_address{namespace=\"vedoc-core-dev0\",endpoint=\"event-brk-dev0-kafka-brokers\",ip=\"172.17.39.61\",ready=\"true\"}"

Now, we have compared the event-brk-<env>-kafka-brokers endpoint on version ES 12 with Endpoint in version ES 11.8 (ES operator version 3.8) and there is visible that inside the endpoint there are duplicate addresses subsets for kafka broker and controller.

We opened a PMR/SupportCase for this and there was some battling between whether ES Operator has to be fixed or OpenShift has to be fixed and as RedHat is limited in possibility to fix the issue, the conslusion was to raise a feature enhancement to ES to avoid the OCP limitation by splitting the service into 2 different ones.