CVO story: persistence for "how long has this ClusterOperator been updating?". For a first pass, we can just hold this in memory in the CVO, and possibly expose it to our managers via the ResourceReconcilliationIssues structure created in OTA-1159. But to persist between CVO restarts, we could have the CVO annotate the ClusterOperator with "here's when I first precreated this ClusterOperator during update $KEY". Update keys are probably (a hash of?) (startTime, targetImage). And the CVO would set the annotation when it pre-created the cluster operator for the release, unless an annotation already existed with the same update key. And the reconciling-mode CVO could clear off the annotations.

We are not sure yet whether we want to do this with CO annotations or OTA-1159 "result of work", we defer this decision to after we implement OTA-1159.

I ended up doing this via the OTA-1159 result-of-work route, and I just did the "expose" side.  If we want to cover persistence between CVO-container-restarts, we'd need a follow-up ticket for that.  The CVO is only likely to restart when machine-config is moving though, and giving that component a bit more time doesn't seem like a terrible thing, so we might not need a follow-up ticket at all.