The etcd cluster is considered failed if the majority of etcd members have permanently failed.

After the etcd cluster failure, all running workload might continue operating. However due to etcd role, Kubernetes cannot make any changes to its current state. Although the scheduled pods might continue to run, no new pods can be scheduled. So applications running in Kubernetes cluster might continue to serve traffic but etcd cluster should be recovered as soon as possible.

You can find more information about etcd's role in Kubernetes in Rancher docs.