Reconciliation requiring StatefulSet recreation results in new PVC, with old data orphaned

[tcbennun]

Let's say I use this YAML to spec my EMQX resource.

apiVersion: apps.emqx.io/v2beta1
kind: EMQX
metadata:
  name: emqx
  namespace: emqx
spec:
  image: emqx/emqx:5.9.0
  coreTemplate:
    spec:
      resources:
        requests:
          cpu: 250m
          memory: 1Gi
      volumeClaimTemplates:
        storageClassName: standard
        resources:
          requests:
            storage: 1Gi
        accessModes:
          - ReadWriteOnce
      replicas: 1

One day, I realise I'm throttling out on CPU. I change cpu: 250m to cpu: 500m, and apply the new config (in my case I'm using Terraform, FWIW).

My EMQX Operator reconciles, and realises it needs to create a new StatefulSet. In doing so, it also creates a brand-new persistent volume claim. At the end of the day, my EMQX cluster is running anew, with none of its persisted data.

I can certainly manually copy the data from the old, orphaned PVC to the new one, but that kind of defies the point, does it not?

I observed this on operator 2.2.29.