Ranch slow down in kubernetes

I’m seeing behavior I don’t understand in ranch 1.6 under kubernetes.

If I send one connection per second to a ranch TCP port the active_connections and all_connections grow slowly. If I leave it running, active_connections will be around 3000 in two hours. I am processing each event so each one should show up in a database. The number of event in the database plus the number of active_connections equals actual number of events. If I stop the event stream the number of active_connections starts to go down and eventually all the events end up in the database if max_connections is set high enough. My development environment is in vagrant and I don’t see the build up of active_connections there. I’ve read through the user documentation for ranch but nothing jumps out as a possible problem.

Thanks for any help you can give.