What is preheat? Does it help with HA of Docker images?

[applejag]

There's some documentation on how to use preheat, but I'm brand new to the project and I'm having a hard time understanding it.

My intuition says that it's probably like telling the peers to just download some images and keep them cached locally. But to what extent?

The use case I'm after is for having high availability of the images in my cluster. In the (albeit rare) situation of upstream image registry not having a specific image anymore (e.g Harbor/Docker Hub is down, or the image was garbage collected because of age) in combination with a Kubernetes node goes down that hosts the only pods that make use of that given image, then I want the other nodes in the Kubernetes cluster to already have the image replicated so they can successfully start up new pods on other nodes.

Can preheat help solve this?
Can I scrape all Docker images in my cluster and tell my cluster to preheat all those images on all my worker nodes?

Or does preheat mean "just make sure the image is available somewhere on the cluster nodes/peers".

Is there any replication or preheat configuration I could set to make sure I would be fine if a node and upstream Docker registry goes down at the same time?