What would you like to be added?

A mechanism to set a Pod into phase=Failed when the image pull has failed for a number of times (perhaps configurable).
Currently, the Pod just stays in phase=Pending

Why is this needed?

This is especially problematic for Jobs submitted through a queueing system.

In a queued environment, the time when the job starts running (pods created) might be hours or even days after the Job is created. Then, the user that submitted the job might not realize their mistake until it's too late. Since these Pods block resources in the cluster, it might cause other pending Jobs not to start.

If the Pods stay in phase=Pending, the job controller cannot do anything about them, as it only does "failure handling" once the Pods actually terminate with a phase=Failed.