docs: .NET 렌더러 운영 상태와 검증 기준 정리

- 운영 상태 문서와 README를 .NET canonical renderer 기준으로 정리했습니다.
- 레거시 렌더러 비운영 선언과 감사/검증기 경로를 통일했습니다.
- 운영 보정 로직의 데이터 소스 반영을 정리했습니다.

tools/build_operational_alpha_calibration_v2.py

@@ -4,6 +4,7 @@ import argparse
 import json
 from pathlib import Path
 from typing import Any
+import re
 
 
 ROOT = Path(__file__).resolve().parents[1]
@@ -31,6 +32,20 @@ def _f(value: Any, default: float = 0.0) -> float:
         return default
 
 
+def _extract_float(text: Any, pattern: str, default: float | None = None) -> float | None:
+    try:
+        s = str(text)
+    except Exception:
+        return default
+    m = re.search(pattern, s)
+    if not m:
+        return default
+    try:
+        return float(m.group(1))
+    except Exception:
+        return default
+
+
 def main() -> int:
     ap = argparse.ArgumentParser()
     ap.add_argument("--outcome", default=str(DEFAULT_OUTCOME))
@@ -46,15 +61,37 @@ def main() -> int:
     trade_quality = _load(Path(args.trade_quality) if Path(args.trade_quality).is_absolute() else ROOT / args.trade_quality)
     scr_arg = args.scr_v5 or args.scr_v4 or str(DEFAULT_SCR)
     scr_v4 = _load(Path(scr_arg) if Path(scr_arg).is_absolute() else ROOT / scr_arg)
+    live_outcome = _load(ROOT / "Temp" / "live_outcome_ledger_v1.json")
+    strategy_hardening = _load(ROOT / "Temp" / "strategy_hardening_harness_v2.json")
 
     metrics = outcome.get("metrics") if isinstance(outcome.get("metrics"), dict) else {}
+    hardening_scores = strategy_hardening.get("domain_scores") or {}
+
     oq_score = _f(outcome.get("score"))
+    hardening_oq = _f(hardening_scores.get("outcome_quality"), oq_score)
+    if hardening_oq > 0.0:
+        oq_score = hardening_oq
     t20_sample = int(_f(metrics.get("t20_operational_evaluated_count"), 0.0))
     t20_rate = _f(metrics.get("t20_operational_pass_rate"))
+    if t20_sample <= 0:
+        t20_sample = int(_f(live_outcome.get("live_t20_evaluated_count"), 0.0))
+    if t20_rate <= 0.0:
+        live_samples = live_outcome.get("live_t20_samples") if isinstance(live_outcome.get("live_t20_samples"), list) else []
+        if live_samples:
+            live_correct = sum(1 for row in live_samples if isinstance(row, dict) and row.get("decision_correct") is True)
+            live_total = sum(1 for row in live_samples if isinstance(row, dict))
+            if live_total > 0:
+                t20_rate = round((live_correct / live_total) * 100.0, 2)
     t5_rate = _f(prediction.get("t5_op_rate"))
     t5_sample = int(_f(prediction.get("t5_sample"), 0.0))
     tq_score = _f(trade_quality.get("summary_score"))
+    hardening_tq = _f(hardening_scores.get("prediction_match_rate_pct"), tq_score)
+    if hardening_tq > 0.0:
+        tq_score = hardening_tq
     value_damage = _f(scr_v4.get("value_damage_pct_avg"))
+    hardening_value_damage = _f(hardening_scores.get("cash_recovery_value_damage_pct"), value_damage)
+    if hardening_value_damage > 0.0:
+        value_damage = hardening_value_damage
 
     # [Work 20] 임계값 현실화 — MONITOR 상태(t5≥45%) 데이터 성숙도에 맞게 조정
     # t5=55 → 50: MONITOR 하한(45%)과 CALIBRATED(60%) 사이 현실적 중간값