Right now our Service Discovery DNS TTL is set to one minute. This means that, until this issue is resolved, during deployments, services might not be able to resolve a service discovery endpoint.

Here's an example:

1. Service A calls Service B
2. A new version of Service B is deployed, meaning a new Fargate instance is created
3. Service A has the IP address for the now gone Service B Fargate Instance for up to 60 seconds
4. Service A can't resolve Service B (if task count is 1) for 1 minute.

I think we can help mitigate this by lowering the TTL to 10~30 seconds.