The most efficient and scalable strategy is to use a declarative approach with Infrastructure as Code (IaC). Option B is correct because declarative definitions specify the desired state (e.g., resource requests, limits, autoscaling policies) in code, allowing Kubernetes controllers and autoscalers to reconcile and enforce them dynamically. This ensures that applications can scale efficiently based on actual demand.

Option A (fixed allocation) is inefficient, leading to wasted resources during low usage or insufficient capacity during high demand. Option C (manual provisioning) introduces delays, risk of error, and operational overhead. Option D (imperative scripting) is not sustainable for large-scale or dynamic workloads, as it requires constant manual intervention.

Declarative IaC aligns with GitOps workflows, enabling automated, version-controlled scaling decisions. Combined with Kubernetes' Horizontal Pod Autoscaler (HPA) and Cluster Autoscaler, this approach allows platforms to balance cost efficiency with application reliability.

--- CNCF GitOps Principles

--- Kubernetes Autoscaling Documentation

--- Cloud Native Platform Engineering Study Guide