ArgoCD 2.11 - Loop of PATCH calls to Application objects

[diranged]

Is it possible that #18061 has suddenly changed the behavior of the Application Controller so that it isn't comparing things the same way? We are seeing a massive (relatively speaking) jump in the PATCH requests/second that the ArgoCD App controller is making now across ~10 clusters that we did this upgrade on:

I suspect the issue has to do with how the code is comparing the valueFiles lists from before and after.. we had some valueFiles lists that had duplicate values.. and when we run kubectl get applications -w -o json | kubectl grep -w, we see diff's like these:

        revision: "794fcbae2671366663e107cd50079c2e96894591"
        source:
          helm:
            ignoreMissingValueFiles: true
            valueFiles:
              - "values.yaml"
              - "values.staging-xx.yaml"
              - "values.staging-xx.yaml"
              - "values/staging/values.yaml"
              - "values/staging/staging-xx/values.yaml"
              - "values/staging/staging-xx/types/values.otherfile.yaml"
              - "values/staging/staging-xx/values.staging-xx.yaml"
-             - 
+             - "values.staging-xx.yaml"
            values: "
cluster:
  accountId: ".."

Originally posted by @diranged in #18061 (comment)

[diranged]

We're reverting back to 2.10.9... but this is happening on every cluster we have across the board. We caught this in staging, so ~10 clusters or so.

[diranged]

We just tried the upgrade again just for fun (no, I actually forgot about the issue and clicked the upgrade button ... 🤦) - and it's reliable, it happens again right away. Any suggestions on what to do to troubleshoot it?