In our setup we're using Karpenter to scale our nodes. Whenever a node gets terminated it could happen that the controller pod is terminated as well and the hook to inject a sidecar would be missed.
To prevent that, we can use PodDisruptionBudgets.
Here's an example from the cert-manager chart: https://artifacthub.io/packages/helm/cert-manager/cert-manager#poddisruptionbudget-enabled-~-bool