Currently, the type: internal listener is using the headless service and the DNs names it gives tot he pods to route the traffic. This has some advantages and disadvantages. In some cases, it could be better to have an internal listener based on per-broker Cluster IP services rather than on the headless service. For example:

• When building manually your own external listener based on Ingress or Gateway API or something else
• When using tools such as Telepresence which have problems routing the headless services