The Retry Policy is based on conventional HTTP responses and isn't specific to this action, but a common behavior for all actions. So, a workaround would be to detect the error and loop over the action as required.
This of course would be difficult if required for multiple actions and/or in multiple locations. If it's just for one action, you could extract the section into a separate workflow that you could reuse.
Also, feel free to raise a feature request to address this difference.