When applying an image override for the GatewayService human errors can sneak in. If there is a mistake in the image path causing it to be invalid the StatefulSet under the GatewayService will have two Running pods and a third in an ErrImagePull state. This is only an issue when the GatewayService has spec.updateStrategy.mode is set to automatic which is the default for API Gateways.

small-ocp-gw-0                                                   1/1     Running        0             23m
small-ocp-gw-1                                                   1/1     Running        0             20m
small-ocp-gw-2                                                   0/1     ErrImagePull   0             7s

The consequence of this is that by default the GatewayService will not apply updates to the StatefulSet while it is in an error state. Therefore correcting the image in the GatewayService objects will not resolve this. In order to recover this without causing an outage, you can modify the GatewayService Object with the following configuration.

spec:
  dataPowerOveride:
    image: imagurl.com/cp/datapower@sha:123123123123123

  updateStrategy:
    mode: manual
    partition: 0

This will allow the stateful set to be updated even if it is in an errored state. I do recommend removing the spec.updateStrategy after the new image is rolled out.