Is it expected that each queue creates a new DB connection independent of the connection pool maintained by Ecto? It is how I see EctoJob behave in our production and it seems very weird to me. So in effect there is pool_size + number_of_queues connections.
