Controlled Load Layer 1 for the Solana Kubernetes Observability Core

This preprint describes Controlled Load Layer 1 for the Solana Containerised Testbed. The work extends a validated KVM-based multi-node Kubernetes observability baseline with controlled, reproducible transaction load generation. The methodological sequence is infrastructure, observability, and then controlled load. The implementation introduces a containerised transaction load generator, CSV collection tooling, Kubernetes manifests, PVC-backed result persistence, and validation procedures for local and in-cluster controlled-load execution. The controlled-load layer runs inside the existing `solana-observability` namespace and uses the in-cluster Solana RPC and metrics Services already validated by the observability baseline. A candidate Kubernetes run using commanded load levels of `1`, `2`, `4`, and `8` transactions per second demonstrates that the controlled-load layer can drive transaction submission through an in-cluster Service, collect telemetry, persist results, and return to a safe idle state. The achieved mean rates were `1.015567`, `1.895051`, `3.822429`, and `7.664078` transactions per second, respectively, with zero observed error rate in the derived per-level summary. After the run, `loadgen2` reported `sent_total=1401`, `ok_total=1401`, `err_total=0`, `inflight=0`, and an empty `last_err`. This stage is intentionally limited to controlled load generation and observability integration. It does not introduce MPC, reinforcement learning, MARL, adaptive policy learning, or Agave migration. The result is a reproducible experimental layer for later research into saturation analysis, control strategies, and high-load behaviour of Solana-oriented Kubernetes testbeds.