I use Docker Swarm + Multi-Host Overlay Driver (with consul KV store) + Docker Compose.
Sometimes, after killing and removing containers with

docker-compose kill
docker-compose rm -v

I cannot start containers again because I get

Cannot start container 0bdb1cfdc4e21c0aaa51a161076e975ed60d2e6d7c66385a870df8b9ab0bac5b: a container has already joined the endpoint

When I start a simple container on any host from my cluster and check /etc/hosts I can see that it contains same entries as before killing and removing containers.
Why is it happening is to investigate (it happens sometimes, I haven't found any pattern yet, but I'll try).
But my question is what can I do in case of not fully removed conteiners from network? Is it possible to do some cleaning manually? Now I can't start my docker-compose containers because it always uses same names for containers and it is conflicting with old names and it makes my cluster pretty unusable...