We're currently using k8s-image-swapper to proxy all the images in our cluster to ECR. I have setup a lifecycle policy for the images to delete after 30 days after push.

However once the image gets pruned by ECR k8s-image-swapper won't replace it if the image is requested again causing an error pulling the image because it no longer exists in ECR. Only fix I've found is to restart k8s-image-swapper pods and then the image gets pulled and pushed to ECR correctly.