ConsolePluginComponents CMO task depends on the availability of a conversion service that is part of the console-operator Pod, that Pod is not duplicated, thus when it restarts due to a cluster upgrade or other that conversion webhook becomes unavailable and all the ConsolePlugin API queries from that CMO task fail.

Create a 4.14 cluster, make the console-operator unmanaged and bring it down, watch the ConsolePluginComponents tasks fail instantly after they're run.

    1.
    2.
    3.
    

The ConsolePluginComponents tasks fail instantly after they're run.

The tasks should be more resilient and retry.

The long term solution is for that ConsolePlugin conversion service to be duplicated.

For OCP >=4.15, CMO v1 ConsolePlugin querie no longer rely on the conversion webhook because of https://github.com/openshift/api/pull/1477.

But the retries will keep the task future proof + we'll be able to backport the fix.