I've been using Docker Swarm in production for about 2 years now... processing about 1TB of data / month across 30+ containers. The networking and routing has been rock solid except for that one day that the Docker dev team, in one release, accidentally added random hashes to the internal DNS names of services. Ever since that day I've used the docker-compose network alias for internal routing https://docs.docker.com/compose/compose-file/#aliases

Discovering bugs in a technology you just started "looking into" actually sounds like the learning curve.