Models pods duplications after corrupted deployment

[OrBarda]

when trying to deploy a model right after deployment of an older version of the model that failed, It seems that Seldon deployment doesn't terminate old pods(probably because they are in fail mode), and it actually creates new Kubernetes deployments.
The only change was the image, and after testing it a little it seems to happen only when the pod fails.