fix(game1): 🐛 update p1-30 status to partial and add detailed batch results

Co-Authored-By: Lilith Autocommit <noreply@atlilith.com>

.project/objectives/p1-30.md

@@ -2,11 +2,11 @@
 id: p1-30
 title: "Optimize `_build_tactical_state` — 8000-tile GDScript dict-build per AI turn blocks p1-22 huge-map gate"
 priority: p1
-status: missing
+status: partial
 scope: game1
 tags: [perf, tactical-ai]
 owner: warcouncil
-updated_at: 2026-04-26
+updated_at: 2026-05-03
 ---
 ## Summary
 
@@ -37,4 +37,12 @@ Option 2 is more correct (Rail-1) but bigger surface. Option 1 is the quick win.
   - mc-ai lib tests: 208/208 pass (`cargo test -p mc-ai --lib`)
   - api-gdext: `cargo check` clean (no errors)
 - ❌ Re-run `tools/huge-map-5clan.sh` 10-seed batch — verify ≥5/10 victories with ≥2 distinct winners (closes p1-22's failing sub-gate)
-- ❌ p1-22's evidence updated to cite this objective's closure
+  - **Batch run 2026-05-03** (`~/.cache/mc-src-20260502_215856/.local/iter/huge-map-5clan-20260503_033046/`):
+    - SEEDS=10, MCTS_DECISION_BUDGET_MS=2000, T500
+    - **Result: 0/10 victories, 0 distinct winners** — verdict.json `pass: false`
+    - All 10 seeds reached the 1800s seed wall-clock timeout while still `outcome=in_progress`
+    - Median game length: turn 243 (seeds spread T100-450)
+    - The Rail-1 tile-state-in-Rust handoff DID land (208/208 tests still pass) — but the strategic-decision wall-clock at 2000ms per AI turn × 5 AI × 100-450 turns × 1800s seed budget is not enough to reach decision in the huge-map config
+  - **Diagnosis**: The tactical-state serialization fix is necessary but insufficient. Either MCTS_DECISION_BUDGET_MS needs to drop (faster decisions, lower quality) or the per-seed wall-clock budget (currently 1800s) needs to grow significantly. Neither is a code change — both are knob retunings or a deeper MCTS efficiency improvement.
+  - **Status remains `partial`** — the implementation work landed; the gate metric did not move past R0 baseline (which was 2/10 victories — this run got 0/10).
+- ❌ p1-22's evidence updated to cite this objective's closure (still ❌ — gate remains failing)