Synced from resonance-engine-active - July 16 2026

test_diff_output.txt

@@ -0,0 +1,47 @@
+Settling for 2600 cycles...
+Settle complete. Last cycle: 34460
+Checkpoint saved at cycle 34461
+Using checkpoint: ckpt_20260715_104712_c34461.bin
+
+=== CONTROL RUN (no perturbation) ===
+Control: 260 frames, cycles 34500 - 37090
+
+Loading checkpoint for Arm A...
+Post-load settle: last cycle 35060
+
+=== ARM A: inject_density ===
+Arm A: 260 frames
+
+Loading checkpoint for Arm B...
+Post-load settle: last cycle 35060
+
+=== ARM B: perturb_stress ===
+Arm B: 260 frames
+
+============================================================
+DIFFERENTIAL HOLD-TIME ANALYSIS
+============================================================
+Baseline noise: 1.724454e-03
+Return threshold (5% of baseline noise): 8.622270e-05
+
+Arm A: inject_density:
+  Distance range: [3.834438e-02, 3.118864e-02]
+  Frames above threshold: 260/260
+  Never returned within window (2600 cycles)
+  Hold-time: 2600 cycles, 95% CI: [2600, 2600]
+
+Arm B: perturb_stress:
+  Distance range: [3.834438e-02, 3.118472e-02]
+  Frames above threshold: 260/260
+  Never returned within window (2600 cycles)
+  Hold-time: 2600 cycles, 95% CI: [2600, 2600]
+
+============================================================
+SANITY GATE: Does density decay faster than stress?
+============================================================
+  Arm A (density): 2600 cycles, CI [2600, 2600]
+  Arm B (stress):  2600 cycles, CI [2600, 2600]
+  GATE FAIL: density (2600) == stress (2600) — instrument STILL broken.
+  STOP. Do NOT proceed to sweep engine until this is fixed.
+
+Done.