I was having problem getting containerd starting after a reboot on a crashed windows node. The message from the containerd said failed to recover state: failed to reserve sandbox name.

I found this issue on github that helped me in the right direction.

Start by modifying config.toml by adding

disabled_plugins = ["io.containerd.grpc.v1.cri"]

Then start containerd and use the ctr cli.

# To list the containers and find the container id from the logs
ctr -n k8s.io containers ls

# Delete the container
ctr -n k8s.io containers rm <containter id>

# Or delete all containers
ctr -n k8s.io containers rm $(ctr -n k8s.io containers ls -q)

Remove the line added to config.toml and restart containerd and everything should work again