fix(magnitude): var_scale permanent floor — eval Full reachable via conviction

Smoke (with mag_concat off-by-one fixed and fxcache regenerated):
- training MAG_DIST: Q=0.287 H=0.301 F=0.412 (healthy diversity)
- INTENT at eval:    Q=0.045 H=0.045 F=0.911 (network learned Full)
- realised at eval:  Q=0.592 H=0.408 F=0.000 (Full silenced)

Even though q_f >> q_h >> q_q (network strongly prefers Full), eval
realised Full = 0.000. Position-sizing pipeline composes four
multiplicative shrinks at experience_kernels.cu:1615-1672:

  effective_max_pos = max_position
                    × cvar_scale            (line 1621)
                    × q_gap_conviction      (line 1628, clamped [0.25,1])
                    × kelly_f               (line 1649, only if >20 trades)
                    × var_scale             (line 1671, = 1/(1+sqrt(var_q)))

Each term well-bounded individually but composing them silences the
policy at validation when var_q persists high (var_q ≈ 30 → var_scale
≈ 0.15; even with conviction = 1.0, kelly_f = 1.0, the compound 0.15
falls in the Quarter bucket [0, 0.375)). Network INTENT reaches the
target_position kernel correctly; var_scale strips it back out.

Same family as val-Flat-collapse (warm-branch Kelly = 0 from balanced
priors) — fix is the same pearl
(`pearl_blend_formulas_must_have_permanent_floor.md`):

  var_scale = max(var_scale, q_gap_conviction)

The q_gap-derived conviction is already an adaptive ISV-coupled signal
(line 1628), already clamped to [0.25, 1.0]. Using it as a permanent
floor on var_scale lets the policy's magnitude intent reach the
realised position when conviction is high, regardless of variance.
No tuned constants — feedback_adaptive_not_tuned + feedback_isv_for_adaptive_bounds
both honoured by reusing the existing adaptive bound rather than
introducing a new threshold.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

crates/ml/src/cuda_pipeline/experience_kernels.cu

@@ -1654,7 +1654,19 @@ extern "C" __global__ void experience_env_step(
     /* Distributional variance position sizing (Kelly from C51 atoms).
      * High Var[Q] = uncertain outcome → smaller position.
      * Low Var[Q]  = confident → full position.
-     * Scale: 1 / (1 + sqrt(Var[Q_taken])). Bounded (0, 1]. */
+     * Scale: 1 / (1 + sqrt(Var[Q_taken])). Bounded (0, 1].
+     *
+     * Permanent floor: when the policy's own conviction signal (q_gap) is
+     * high, var_scale must not silence the magnitude intent. Same pearl as
+     * the Kelly warmup_floor (`pearl_blend_formulas_must_have_permanent_floor.md`):
+     * cascading multiplicative shrinks (cvar × q_gap_conviction × kelly × var_scale)
+     * compose to near-zero at validation when var_q persists high, even
+     * though every individual term is "well-bounded". The fix is the same:
+     * floor var_scale at the q_gap-derived conviction already applied
+     * upstream — it IS the adaptive bound (per feedback_isv_for_adaptive_bounds.md),
+     * not a tuned constant. With this floor, high conviction overrides
+     * variance shrinkage so Full intent at eval can produce Full realized
+     * positions even before var_q has fully decayed. */
     if (q_variance != NULL) {
         /* Taken action's factored index: sum of branch-local indices */
         int taken_branch_idx = dir_idx;  /* direction is branch 0 */
@@ -1664,6 +1676,11 @@ extern "C" __global__ void experience_env_step(
         /* Use max variance across the two exposure branches (direction + magnitude) */
         float var_q = fmaxf(var_taken, var_mag);
         float var_scale = 1.0f / (1.0f + sqrtf(var_q));
+        /* Permanent floor — same q_gap-derived conviction clamped at line 1628. */
+        float var_floor = (q_gaps != NULL)
+            ? fminf(fmaxf(q_gaps[i] * 0.5f, 0.25f), 1.0f)
+            : 0.25f;
+        var_scale = fmaxf(var_scale, var_floor);
         /* Task 0.3: persist the per-sample Kelly scale for host-side epoch-mean
          * reduction. Host filters on `> 0.0f` to distinguish measured samples
          * from unreached/Var[Q]-disabled default-zero slots. */

docs/dqn-wire-up-audit.md

@@ -2,6 +2,20 @@
 
 **Status:** Populated during Plan 1 Task 6 (A.5 orphan audit). Updated on every commit per Invariant 7.
 
+var_scale permanent floor for eval magnitude collapse (2026-04-27): the
+EVAL_DIST=Quarter 0.59 / Half 0.41 / Full 0.00 pathology persists even with
+intent=0.911 for Full because the position-sizing pipeline composes four
+multiplicative shrinks (cvar_scale × q_gap_conviction × kelly_f × var_scale)
+each well-bounded individually but compounding to ≪ 0.75. Kelly already
+fixed via permanent-floor pearl (`pearl_blend_formulas_must_have_permanent_floor.md`);
+applied the same pearl to var_scale at `experience_kernels.cu:1666`:
+`var_scale = max(var_scale, q_gap_conviction)`. The conviction signal IS
+the adaptive bound (per `feedback_isv_for_adaptive_bounds.md`), already
+clamped to [0.25, 1.0] at line 1628. With this floor, high conviction
+overrides Var[Q] shrinkage so the policy's magnitude intent reaches the
+realised position even before var_q decays. No tuned constants — same
+adaptive q_gap signal used upstream.
+
 data_source globalization to mbp10 (2026-04-27): completes the alignment
 started in the earlier "fxcache data_source alignment" entry below. Smoke
 (`dqn-smoketest.toml`) and localdev (`dqn-localdev.toml`) profiles