Phase-1 pilot — machinery-blocked partial (no runs completed)

Outcome: all three success criteria UNRESOLVED. Zero simulation cells executed; the session produced a corrective machinery finding, not data.

What happened. The initial batch fanned out one sim-runner per staged floor (10 floors × seeds [0,1,2] × occupancy grid n_agents∈{0,2,4,6,8} @ 2.4 GHz = the 150-cell Phase-1 design). All 10 launches failed fast and identically at config validation: exp-csi-static's StaticParamsSchema rejects top-level grid and seeds as extra_forbidden. The campaign's sim_params block is a sweep specification (floors×grid×seeds), not a single-run param dict — it cannot be handed to sim_run_launch. No compute or PostGIS staging occurred (failures at ~0.05 s), so the :5435 prerequisite remains unverified.

Criterion status. (1) Primary inflation-ratio CI — unresolved (was pre-declared under-powered at n=10 regardless). (2) Variance split σ²_floor vs σ²_seed — unresolved, no per-floor CV(N) produced. (3) Comparator honesty — untestable without a LOFO reduction.

Actionable fix for the follow-on session. Re-dispatch the identical design through sim_sweep_launch (dry_run→launch), which expands the replications×grid×seeds axes natively and attaches per-cell runs to the session; then sim_reduction_run (multi_seed_ci + variance components) over the cross_geometry.parquet corpus, and render the mandated csi_cross_geometry small-multiples + seed_trace figures. Verify PostGIS :5435 reachability from the runner first (still a live risk per the 2026-07-06 sibling failure). Budget was exhausted this session before the corrected sweep could run.