Out-of-memory on the consul-dataplane ecs task don't force the task replacement #325
Description
In v0.8.x, if a ECS service runs close to the allocated memory limit, the ECS orchestrator might decide to stop one of the consul sidecars.
When this happens, given none of consul sidecars are marked as essential (see mesh-task and gateway-task), the service will become unreachable from consul mesh but the task wouldn't be replaced until a manual redeployment happens.
We can ammend this on our fork, but it's a bit annoying to replicate all these fixes every new release. Can you consider to have this fixed on the origin?
