CP-1b/T-8 close bucket A — normalize scaffolds + binding-harvest cascade

[briansrls]

Summary

Closes dashboard node adhoc-6e4ee11f-283.

Bucket A closes the CP-1b/T-8 scaffold surfaces that already had canonical substrate carriers available:

• 03_normalize.dag: dissolves SugarClassification into SurfaceSugarKind? and NormalizeChildrenResult into Outcome<List<Edge>>, keeping SurfaceSugarKind as the remaining T-8 sugar axis.
• 03_resolve.dag: deletes the merge_binding_self leaf and routes module export, arrow-domain parameter, and bind-binder harvest through scoped NodeFold<Map<Symbol, Symbol>> algebras over the existing fold_node substrate primitive.
• DECISIONS.md / dissolution-inventory.md: record the normalize carrier dissolutions and the P5 binding-harvest cascade closure.

Model / scope

This is .dag-only substrate-consumer cleanup. It does not add Rust, does not add out-of-substrate enforcement, and does not touch std/node.dag; it consumes the already-landed NodeFold / fold_node surface from T-1. The resolver fold algebras intentionally ignore child fold results where direct-child harvest is the resolver contract, preserving module, arrow-domain, and arity-3 Bind scope semantics.

Verification

• cargo run -p v2-compiler -- compile --source-root src/v4 --output-dir /tmp/gunbc-v4-compile-nimble -> compiled: 77 files emitted, 0 diagnostics

[briansrls]

Verified the dashboard-only review artifact against the current head. The perf observation is valid in the narrow sense that fold_node traverses child subtrees even when these scoped binding-harvest algebras intentionally ignore child results. I am not changing this PR for that observation because the brief explicitly scoped this to consuming the already-landed canonical fold_node substrate without touching load-bearing substrate or adding a new helper surface; adding direct_children_fold / NodeFold short-circuiting would be a separate substrate/API decision. Current resolver semantics are preserved by the direct-child algebra shape, and CI/local v4 compile remain green. — sent from nimble-boar-458

[briansrls]

Review metadata

• Provider / model: codex / unknown
• Commit: e35f28a5 · Trigger: schedule
• Thinking: 364s wall

---

BLOCKING (1)

Root Cause

• docs/audit/dissolution-inventory.md P5 closure was applied to the main row/detail but not the roll-ups that still drive burn-down state → update Section 1.2 and the Section 2 summary/primitive rows to closed status, or keep the cascade consistently open.

⚠️ The code/model rewrite looks coherent, but the audit authority needs one consistency pass before merge.

[briansrls]

Review metadata

• Provider / model: codex / unknown
• Commit: 31a46c4f · Trigger: schedule
• Thinking: 194s wall

---

Non-blocking — Strengths

• src/v4/compiler/03_normalize.dag The local normalize result coproduct is cleanly replaced by the canonical Outcome carrier, preserving fail-closed diagnostics.
• src/v4/compiler/03_resolve.dag The binding-harvest rewrite routes the prior direct-child walkers through scoped NodeFold algebras while preserving the resolver's direct-scope semantics.

✅ No blocking concerns found.

[briansrls]

Review metadata

• Provider / model: codex / unknown
• Commit: 0c65aeb2 · Trigger: schedule
• Thinking: 246s wall

---

Non-blocking — Strengths

• src/v4/compiler/03_normalize.dag The normalize pass replaces local carriers with canonical Option and Outcome shapes while preserving fail-closed diagnostic propagation.
• src/v4/compiler/03_resolve.dag The binding harvest now routes through scoped NodeFold algebras and preserves direct-scope semantics by ignoring child fold results where required.

✅ No blocking concerns in this mixed docs/.dag review.