The defining capability of GitOps controllers such as Argo CD and Flux is continuous reconciliation: they compare the desired state stored in Git to the actual state in the cluster and then alert and/or correct drift, making D correct. In GitOps, Git becomes the single source of truth for declarative configuration (Kubernetes manifests, Helm charts, Kustomize overlays). The controller watches Git for changes and applies them, and it also watches the cluster for divergence.
This is more than "auditing human changes" (option A). GitOps does provide auditability because changes are made via commits and pull requests, but the core functionality is the reconciliation loop that keeps cluster state aligned with Git, including optional automated sync/remediation. Option B is not accurate because GitOps is not about tracking user Git commands; it's about reconciling desired state definitions. Option C (automatically creating pull requests for outdated dependencies) is a useful feature in some tooling ecosystems, but it is not the central defining behavior of GitOps controllers.
In Kubernetes delivery terms, this approach improves reliability: rollouts become repeatable, configuration drift is detected, and recovery is simpler (reapply known-good state from Git). It also supports separation of duties: platform teams can control policies and base layers, while app teams propose changes via PRs.
So the verified statement is: GitOps tools continuously reconcile Git desired state with cluster actual state-exactly option D.
________________________________________