TCP connecter does not check for DNS TTL expiration

[jean-airoldie]

See this issue for more info.

[jean-airoldie]

For now we should stick to using static IPs for known endpoints.

[jean-airoldie]

This can be a major issue if a bound socket is used in a docker swarm or k8 cluster since they use their owned DNS to refer to containers.

[jean-airoldie]

Getting the DNS TTL in a cross platform manner is probably quite a pain, to as a temporary mitigation, a static DNS_TTL could be set.

[jean-airoldie]

Another temporary client work around would be to call disconnect then connect on last_endpoint again after the recv_timeout period expired.
edit: This is event better than a TTL since we can detect the DNS update as soon as it occurs.