In Kubernetes, it is standard practice (and often a requirement) to associate a StatefulSet with a Headless Service to provide stable network identities. My current understanding of the mechanism is as follows:

1. A Headless Service (clusterIP: None) does not provide a Virtual IP (VIP) or perform traditional load balancing.

2. Each Pod in the StatefulSet is assigned a stable FQDN (e.g., pod-0.service-name.namespace.svc.cluster.local).

3. When resolving these FQDNs, CoreDNS looks up the Service first and then retrieves the individual Pod IP from the associated EndpointSlice.

My Question: Technically, the Pod IP information already exists within the EndpointSlice object. Why is the Service object strictly necessary as an intermediary for DNS resolution?

Is there a specific design principle or architectural reason why CoreDNS doesn't query the EndpointSlice directly for Pod identities? Does the Service perform other critical functions for StatefulSets that I might be overlooking?