IngressRouteTCP always routes to same pod (Load balancing doesn't work)

Traefik Traefik v2
,
1

I created a simple IngressRouteTCP and a service. The service is a regular ClusterIP - points to two Pods (scaled deployment). I do see two Pods in Traefik dashboard so it seems traefik detect that but it always routes to the same Pod. When trying to setup a regular loadbalancer with external IP (and not use traefik) everything works.

Version 2.1.4 (but tested also on 2.1.9)

apiVersion: traefik.containo.us/v1alpha1
kind: IngressRouteTCP
metadata:
  name: ingressroutetcp-livestream
spec:
  entryPoints:
    - rtmp
  routes:
  - match: HostSNI(`*`)
    services:
    - name: livestream-service
      port: 1935
      TerminationDelay: 400

Service:

....
  type: ClusterIP
  ports:
    - port: 1935
      targetPort: 1935
      protocol: TCP
      name: rtmp

My traefik TOML:

# traefik.toml
[log]
  level = "DEBUG"
[serversTransport]
  insecureSkipVerify = true
[ping]
[api]
  dashboard = true
  insecure = true
[entryPoints]
  [entryPoints.traefik]
    address = ":9000"
  [entryPoints.web]
    address = ":8000"
  [entryPoints.websecure]
    address = ":8443"
  [entryPoints.rtmp]
    address = ":1935"
[providers.kubernetesCRD]
[providers.kubernetesIngress]
[[tls.certificates]]
  certFile = "/ssl-dev/tls.crt"
  keyFile = "/ssl-dev/tls.key"

Full issue/config here: https://github.com/containous/traefik/issues/6566