how to handle the possibility that the agent hasn't started yet?

[adamdavis40208]

We're noticing that sometimes a new k8s node will get created, a node service placed on the cluster, and that pod spazzing out because the agent isn't available.

DATADOG TRACER DIAGNOSTIC - Agent Error: socket hang up
Received SIGTERM. Shutting down
SIGINT Received, stopping consumer

Sure, we just spin up another pod and try again, but it'd be nice if this could be retried a few times

Any ideas?