fix(rl): tighten V envelope bootstrap ±200 → ±10 (post-scale calibration)

Phase 2.0 seeded the V envelope at ±200 reasoning from dd049d9a4 dollar
magnitudes ($211 avg win, $235 avg loss). But the envelope clamps the
POST-SCALE V regression target — after apply_reward_scale divides by
mean_abs_pnl_ema to keep typical reward magnitudes ~1.0, the target's
natural range is `r + γ × v_tp1 ≤ ~3` in atom-support units, NOT
dollars. A ±200 bootstrap was a no-op (envelope wider than any possible
target excursion) until the EMA tightened it.

±10 is the tighter bootstrap matching the actual post-scale operating
range while still wide enough to admit normal target excursions. The
kernel's existing sentinel-zero bootstrap discipline at line 98
(pearl_first_observation_bootstrap) snaps the EMA to the first observed
trade magnitude as soon as a done event fires, so this bootstrap only
matters for the first ~10-20 steps.

Note: this commit does NOT resolve the step-4 NaN that surfaced after
the atomicAdd determinization (10d4614fb) — l_v is identical 6.329 ±
0.001 across all ±10 / ±200 / pre-fix runs, proving V predictions are
not the proximate NaN cause. The tightening is still correct per
codebase discipline (pearl_controller_anchors_isv_driven +
feedback_isv_for_adaptive_bounds) and removes one masked confound for
the next debug pass. Suspected actual NaN cause shifted to advantage
RMS / encoder h_t / shared streaming statistics — see
pearl_atomicadd_masks_v_instability for the open investigation.

Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>

crates/ml-alpha/cuda/rl_v_target_envelope_update.cu

@@ -111,9 +111,12 @@ extern "C" __global__ void rl_v_target_envelope_update(
     // Unconditional write: v_head_bwd consumes 594/595 every step. When
     // the EMA is still at sentinel zero (no done events observed yet),
     // we DON'T overwrite the envelope — the trainer-seeded bootstrap
-    // values (±$200 per spec §3 sub-phase 2.0) remain in effect. This
-    // makes the envelope a no-op until the EMA captures a real
-    // trade-magnitude signal, after which it tightens to ±k × ema.
+    // values (±10 post-scale units, tightened from ±200 on 2026-05-29
+    // after the bootstrap was found to be wider than the actual
+    // post-scale target range; see integrated.rs near RL_V_TARGET_MAX
+    // for rationale) remain in effect. This makes the envelope a tight
+    // initial bound until the EMA captures a real trade-magnitude
+    // signal, after which it tightens further to ±k × ema.
     const float ema_now = isv[RL_V_MAGNITUDE_EMA_INDEX];
     if (ema_now > 0.0f) {
         const float k = isv[RL_V_TARGET_K_INDEX];

crates/ml-alpha/src/trainer/integrated.rs

@@ -3112,16 +3112,24 @@ impl IntegratedTrainer {
                 // Phase 2.0 (2026-05-28) — V regression target envelope.
                 // Seeded permissive (±$200 in scaled units; at default
                 // reward_scale=1.0 this matches dd049d9a4's $211 avg win
-                // / $235 avg loss tail). Cold start: envelope is a no-op
-                // — target's natural range is ~[-3, +1] (post-clamp
-                // reward) + γ × v_tp1 (atom span). After done events,
-                // `rl_v_target_envelope_update` tightens the envelope
-                // to ±k × magnitude_ema. k=3 covers ~99% of trade
-                // outcomes. Spec:
+                // 2026-05-29: tightened bootstrap ±200 → ±10. The original
+                // ±200 was a dollar-scale guess from dd049d9a4 ($211 avg
+                // win / $235 avg loss) — but the envelope clamps the
+                // POST-SCALE V target (after apply_reward_scale divides
+                // by mean_abs_pnl_ema so typical reward magnitudes are
+                // ~1.0). In post-scale units a ±10 envelope is already
+                // wide enough to admit normal target excursions
+                // (r + γ × v_tp1 ≤ ~3) while bounding the (v_pred -
+                // target_clamped)² gradient before the EMA takes over.
+                // The kernel's sentinel-zero bootstrap discipline
+                // (rl_v_target_envelope_update.cu:98) snaps the EMA to
+                // the first observed trade magnitude as soon as a done
+                // event fires, so this bootstrap is only in effect for
+                // the first ~10-20 steps. Spec:
                 // docs/superpowers/specs/2026-05-28-state-conditional-q-synthesis.md
                 // §3 Sub-phase 2.0.
-                (crate::rl::isv_slots::RL_V_TARGET_MAX_INDEX,            200.0),
-                (crate::rl::isv_slots::RL_V_TARGET_MIN_INDEX,           -200.0),
+                (crate::rl::isv_slots::RL_V_TARGET_MAX_INDEX,             10.0),
+                (crate::rl::isv_slots::RL_V_TARGET_MIN_INDEX,            -10.0),
                 (crate::rl::isv_slots::RL_V_TARGET_K_INDEX,                3.0),
             ];
             for (slot, value) in isv_constants.iter() {