stream issue with multiple control plane

[jp-gouin]

Environmental Info:
Host Cluster Version:
2 nodes downstream clusters with RKE2 v1.33.5+rke2r1

K3K Cluster Configuration:
hcp mode virtual cluster with 3 control planes

apiVersion: k3k.io/v1beta1
kind: Cluster
metadata:
  name: hcp-server
  namespace: k3k-hcp-server
spec:
  mode: hcp
  tlsSANs:
    - "x.x.x.x"
  servers: 3
  version: v1.33.1-k3s1
  expose:
    nodePort: {}

or
k3kcli cluster create --mode hcp --servers 3 --tls-sans x.x.x.x hcp-server

Describe the bug:
Only one control plane is "active" and handle correctly stream actions (kubectl logs, kubectl exec) the other 2 fail with

time="2026-07-08T12:55:14Z" level=error msg="Sending HTTP/1.1 502 response to 127.0.0.1:54230: dial tcp 10.52.0.5:10250: connect: connection timed out"
time="2026-07-08T12:55:14Z" level=error msg="Sending HTTP/1.1 502 response to 127.0.0.1:54216: dial tcp 10.52.0.5:10250: connect: connection timed out"
time="2026-07-08T12:55:14Z" level=error msg="Sending HTTP/1.1 502 response to 127.0.0.1:54218: dial tcp 10.52.0.5:10250: connect: connection timed out"
time="2026-07-08T12:55:14Z" level=error msg="Sending HTTP/1.1 502 response to 127.0.0.1:54240: dial tcp 10.52.0.5:10250: connect: connection timed out"
time="2026-07-08T12:55:14Z" level=error msg="Sending HTTP/1.1 502 response to 127.0.0.1:54256: dial tcp 10.52.0.5:10250: connect: connection timed out"

And the user gets (2/3 of the time)

kubectl logs -n kube-system metrics-server-7bfffcd44-24kvx
Error from server: Get "https://x.x.x.x:10250/containerLogs/kube-system/metrics-server-7bfffcd44-24kvx/metrics-server": proxy error from 127.0.0.1:6443 while dialing x.x.x.x:10250, code 502: 502 Bad Gateway