Looks like pretix doesn't have graceful shutdowns
We somehow need to do the same thing we have done for the other services, the healthcheck needs to start failing when the task is being scaled down so traefik can remove it from the list of active backends