fix: harden drift verifier and validation flow

.github/copilot-instructions.md

@@ -1040,16 +1040,15 @@ Frequency penalties (overtrading / flip-flop / alternating) now ignore 1-minute
 - Penalty for recent losing trades, bonus for winning streaks
 - **Note:** Analytics check is advisory only - manual trades execute even if rejected by analytics
 
-**Smart Validation Queue (Dec 7, 2025 - TIMEOUT EXTENDED):**
-- **Purpose:** Monitor blocked signals for 30 minutes to confirm price moves
-- **Timeout:** 30 minutes (extended from 10 min based on data analysis)
-- **Rationale:** Analysis of 10 blocked signals showed 30% hit TP1, most moves develop after 15-30 minutes
-- **Example:** Quality 70 signal (ADX 29.7) hit TP1 at 0.41% after 30+ minutes ($22 profit missed with 10-min timeout)
-- **Protection:** -0.4% drawdown limit prevents holding bad signals too long
-- **Configuration:** `entryWindowMinutes: 30` in smart-validation-queue.ts
-- **Trade-off:** Slightly longer hold on losing signals, but data shows most profitable moves take 15-30 min to develop
+**Smart Validation Queue (Dec 10, 2025 - TWO-STAGE CONFIRMATION):**
+- **Purpose:** Monitor blocked signals to confirm price moves before execution (two-stage confirm)
+- **Timeout:** 90 minutes (was 30 minutes) with 30-second checks; restores last 90 minutes on startup
+- **Confirmation:** +0.15% move in trade direction triggers execution; abandon at -0.4% against (unchanged)
+- **Rationale:** Blocked-signal analysis showed rapid +0.15% confirms capture TP1/TP2 while retaining low false positives
+- **Configuration:** `entryWindowMinutes: 90`, `confirmationThreshold: 0.15`, `maxDrawdown: -0.4` in smart-validation-queue.ts
+- **Trade-off:** Longer watch window aligned to two-stage approach; still bounded by drawdown guard
 - **Implementation:** lib/trading/smart-validation-queue.ts line 105
-- **Status:** ✅ DEPLOYED Dec 7, 2025 10:30 CET (commit c9c987a)
+- **Status:** ✅ UPDATED Dec 10, 2025 15:00 CET (two-stage thresholds live)
 
 ## 🧪 Test Infrastructure (Dec 5, 2025 - PR #2)
 
@@ -1062,17 +1061,22 @@ tests/
 ├── helpers/
 │   └── trade-factory.ts        # Factory functions for mock trades
 └── integration/
-    └── position-manager/
-        ├── tp1-detection.test.ts      # 16 tests - TP1 triggers for LONG/SHORT
-        ├── breakeven-sl.test.ts       # 14 tests - SL moves to entry after TP1
-        ├── adx-runner-sl.test.ts      # 18 tests - ADX-based runner SL tiers
-        ├── trailing-stop.test.ts      # 16 tests - ATR-based trailing with peak tracking
-        ├── edge-cases.test.ts         # 15 tests - Token vs USD, phantom detection
-        ├── price-verification.test.ts # 18 tests - Size AND price verification
-        └── decision-helpers.test.ts   # 16 tests - shouldStopLoss, shouldTakeProfit1/2
+  ├── drift-state-verifier/
+  │   ├── position-verification.test.ts   # Identity verification, fail-open bias, cooldown persistence (Bug #82)
+  │   └── cooldown-enforcement.test.ts    # Retry cooldown enforcement and logging (Bug #80)
+  └── position-manager/
+    ├── tp1-detection.test.ts           # 16 tests - TP1 triggers for LONG/SHORT
+    ├── breakeven-sl.test.ts            # 14 tests - SL moves to entry after TP1
+    ├── adx-runner-sl.test.ts           # 18 tests - ADX-based runner SL tiers
+    ├── trailing-stop.test.ts           # 16 tests - ATR-based trailing with peak tracking
+    ├── edge-cases.test.ts              # 15 tests - Token vs USD, phantom detection
+    ├── price-verification.test.ts      # 18 tests - Size AND price verification
+    ├── decision-helpers.test.ts        # 16 tests - shouldStopLoss, shouldTakeProfit1/2
+    ├── monitoring-verification.test.ts # Ensures monitoring actually starts and price updates trigger checks
+    └── pure-runner-profit-widening.test.ts # Profit-based trailing widening after TP2
 ```
 
-**Total:** 7 test files, 113 tests
+**Total:** 13 test files, 162 tests (full suite green as of Dec 10, 2025 before rebuild)
 
 **Test Configuration:**
 - **Framework:** Jest + ts-jest
@@ -2293,6 +2297,7 @@ On container startup, cross-checks last 24h of "closed" trades against actual Dr
 - Impact: Position completely unprotected from downside
 - Detection: Health monitor checks slOrderTx/softStopOrderTx/hardStopOrderTx every 30s
 - Fix required: Validate signatures.length before returning, add error handling around SL placement
+- Additional guard (Dec 10, 2025): tp2SizePercent of 0 or undefined now normalizes to 100% of remaining size so TP2 orders are placed and validation counts stay aligned with expected signatures.
 
 **CRITICAL: Transaction Confirmation Pattern**
 Both `openPosition()` and `closePosition()` MUST confirm transactions on-chain: