I have a setup where Traefik is running behind corporate PLB in Kubernetes, but then it handles all traffic from there.
For a services that is failing, we run 3 instances
