Proposal

KEDA's ScaledObject currently reports Ready=True even when its managed HPA cannot fetch metrics (HPA shows <unknown> targets with ScalingActive=False).
Users see a healthy-looking ScaledObject while scaling is silently broken, the only way to discover the problem is to manually inspect the HPA. Examples of this might be broken metrics adapter, broken APIService.apiregistration, or invalid certs for the gRPC connection between operator and metrics adapter.

This has been reported a few times in the past but imho never properly addressed:

• Keda doesn't fallback to configured replica counts when metrics adapter unhealthy #6475
  • here the fallback would still not be applied but at least user will know that something is wrong and scaling is not happening
• ScaledObject ready condition changed will not fix HPA in some situation #3096
  • partially addressed for from HPA perspective, but the information that something is wrong is still not available on the SO level

In the past, there were issues with too frequent reconciles due to HPA churn

• ScaledObject reconciliation delay due to redundant reconciles triggered by HPA status updates #5281
• Frequent Reconciliation Caused by Minor Changes in HPA Status from Prometheus Scalers #6442

But adding the HPA status aggregation to SO through the scaleHandler scaling loop shouldn't reintroduce any negative performance issues of HPA churn because it won't be based on informer events.

Use-Case

Better UX when SO is not scaling but it's not keda-operator's fault (or is and it just doesn't know, e.g. fails to refresh the certs for metric adapter gRPC).

Is this a feature you are interested in implementing yourself?

Yes

Anything else?

this will benefit from #7579 refactored status updates, ensuring that HPA status aggregation doesn't add unnecessary kube-apiserver load.