Port 8090 hangs

I am able to reproduce very strange scenario.

I have statefulset with 2 varnish pods. I'm sending in a loop purge requests to port 8090 to first Varnish pod. First varnish is properly broadcasting request to second Varnish pod.

While purge requests are flowing, I'm deleting 2nd Varnish pod. First Varnish pod discovers it and removes 1 frontend from the list. I'm repeating this delete 2-3 times in a row.

After few seconds first Varnish node is not accepting calls on port 8090 anymore - cache purge request and curl on port 8090 hangs (from another pod and also from first Varnish pod). I can still connect on port 8080.

Situation never recovers at this point of time and first Varnish is not printing any logs during this time.

Reproduce the issue with two Varnish pods by sending repeated purge requests to port 8090 while deleting the peer pod multiple times. Compare the behavior of ports 8090 and 8080 and inspect the first pod's logs; done means port 8090 continues accepting requests after peer removal and the failure is diagnosable.