Make the branch's large-input story line up with what is actually runnable today. This adds a batch runner for the one-shot stress collector, a structural preflight that checks the documented study surface before new collection starts, and the study docs that explain what is complete on this branch versus what still belongs to the longer draft. The important behavior change is not a new benchmark result. It is that the branch no longer implies a broader archive than it actually ships. The methods, results, and roadmap now say plainly that this branch implements the standard comparison flow, the deterministic round-robin check, the one-shot large-input lane, and the preflight inventory, while session and mitata lanes are still absent here. The stress matrix tooling also makes the archive plan concrete instead of hand-written. It can sweep the ten documented scenario families, default `1 GiB` runs to the cheaper streaming-only profile, and keep the checked-in `16 MiB` smoke artifacts as the lightweight proof that the path works end to end. Verification: - `mise x deno@latest -- deno check experiments/event-shape-study/tools/run_large_input_stress_matrix.ts experiments/event-shape-study/tools/validate_experiment_matrix.ts` - `mise x deno@latest -- deno run --no-lock --allow-read experiments/event-shape-study/tools/validate_experiment_matrix.ts`