Restarting the configurable-http-proxy causes error due to state mismatch

[cloutiertyler]

If the proxy loses it's routing table it can cause errors when spawning containers.

For example, let's say I create an ad-hoc container with the user "username". Some time later the routing table is lost (e.g. because of a new deploy). If I try to use that tmpnb, the proxy will route me to spawn another ad-hoc container because it's not in the table. However, since it was created before it still is alive in Docker, meaning that tmpnb will fail to spawn the container and will return a 500 response.