fix(fxcache): track precompute_features.rs in FEATURE_SCHEMA_HASH

build.rs::emit_feature_schema_hash only hashed
src/features/extraction.rs, src/fxcache.rs, and
../ml-core/src/state_layout.rs. The z-score normalization step lives
in examples/precompute_features.rs:625-631 (added 2026-04-03 in
9f7c14978f) and was NOT covered by the hash. A fxcache written by an
older precompute build (raw features, no normalization) silently
passed today's validate() because every other field matched.

Empirical impact (L40S Argo train-f8h6q, 2026-04-27):

  - PVC fxcache: stale, written pre-normalization → feature column 0
    contains RAW CLOSE PRICES (~$5180 ES futures) instead of z-
    normalized log-returns
  - aux head reads next_states[:, 0] as its next-bar regression
    label (gpu_dqn_trainer.rs:7758-7789)
  - EMA label_scale climbed to 5420 (vs smoke 0.05) → shared trunk
    learned to predict next-bar prices → policy effectively traded
    with future-bar information
  - epoch-0 Sharpe = 141.99 with 0.32% max-drawdown over 214k bars
    — physically impossible; clear future-leak signature

Fix adds examples/precompute_features.rs to schema_sources. New hash
invalidates the stale PVC cache. Argo's ensure-fxcache step has a
regenerate-on-failure branch (infra/k8s/argo/train-template.yaml:
372-383) that auto-regens with current normalized precompute.

Generalises beyond this incident: any future change to feature
normalization, target ordering, or precompute post-processing now
bumps the hash and forces fxcache regen.

Audit entry updated.

crates/ml/build.rs

@@ -281,9 +281,24 @@ fn try_compile_kernel(
 fn emit_feature_schema_hash() {
     // Source files whose bytes determine the on-disk schema.
     // Touching ANY of these files implies "fxcache regen on next deploy".
+    //
+    // 2026-04-27: added `examples/precompute_features.rs` after diagnosing the
+    // L40S `label_scale=5420` (raw-price magnitude) vs smoke `label_scale=0.05`
+    // (z-score) divergence. Root cause: the z-score normalisation step landed
+    // in `precompute_features.rs:625-630` on 2026-04-03 but that file was not
+    // tracked here, so the hash didn't change. PVC fxcaches written by an
+    // earlier build (no normalisation) still validated against the current
+    // binary, silently feeding raw features through column 0 of next_states
+    // and leaking future-bar prices into the policy via the aux head's
+    // shared trunk → impossible Sharpe (141 at epoch 0, 0.32% max DD over
+    // 214k bars). Adding the precompute pipeline to the hash forces every
+    // fxcache to be regenerated when its writer changes — including any
+    // future tweaks to feature normalisation, target ordering, or per-feature
+    // post-processing.
     let schema_sources: &[&str] = &[
         "src/features/extraction.rs",
         "src/fxcache.rs",
+        "examples/precompute_features.rs",
         "../ml-core/src/state_layout.rs",
     ];
 

docs/dqn-wire-up-audit.md

@@ -2,6 +2,28 @@
 
 **Status:** Populated during Plan 1 Task 6 (A.5 orphan audit). Updated on every commit per Invariant 7.
 
+fxcache schema-hash gap fix (2026-04-27): `build.rs::emit_feature_schema_hash`
+hashes only `src/features/extraction.rs`, `src/fxcache.rs`, and
+`../ml-core/src/state_layout.rs`. The z-score normalization step lives in
+`examples/precompute_features.rs:625-631` (added 2026-04-03 in commit
+`9f7c14978f`) and was NOT covered by the hash, so a fxcache written by an
+older precompute build (raw features, no normalization) silently passed
+today's `validate()` because every other field matched. The PVC fxcache
+on the L40S Argo path was such a stale artefact: feature column 0 stored
+RAW CLOSE PRICES (~$5180 ES futures level) instead of z-normalized log-
+returns. The aux next-bar head reads `next_states[:, 0]` as its label
+(`gpu_dqn_trainer.rs:7758-7789`); with raw-price labels, the EMA
+`label_scale` reached 5420 vs smoke's 0.05. The shared trunk learned
+to predict next-bar prices, leaking future info into the policy →
+`train-f8h6q` epoch-0 reported impossible Sharpe=141.99 with 0.32%
+max-drawdown over 214k bars. Fix: add `examples/precompute_features.rs`
+to `schema_sources`. New hash invalidates the stale PVC cache; Argo's
+`ensure-fxcache` regenerate-on-failure branch
+(`infra/k8s/argo/train-template.yaml:372-383`) auto-regens with current
+normalized precompute. Generalises beyond this incident: any future
+change to feature normalization, target ordering, or precompute
+post-processing is now hash-tracked.
+
 mag_stats wr_h/wr_f attribution fix (2026-04-27): `experience_kernels.cu`
 line 1916 binned `action_mag_per_sample` by `actual_mag_core` at every step,
 including trade-close events. `unified_env_step_core` forces