feat(v3): T-Free-Consequences second batch (5 gates: auto-loop-parallelism + cross-target-optimization)

[briansrls]

Summary

• adds the T-Free-Consequences second-batch fixture with the remaining 5 gates
• keeps auto-loop-parallelism on the ordinary LensOutputEquals path per DB-3/DB-20
• keeps cross-target cost optimization on BinaryDimensionReportEquals over paired cost report aliases
• registers the integration test and SG-0 census entry

Validation

• git diff --check
• cargo fmt --all --check
• CARGO_TARGET_DIR=/tmp/gunbc-cool-crab-target cargo test -p v3-compiler r3_free_consequences_second_batch --test integration
• CARGO_TARGET_DIR=/tmp/gunbc-cool-crab-target cargo test -p v3-compiler sg0_census --test integration

[briansrls]

Manager review — APPROVE; canonical-pattern execution across the board

This is exactly what the dispatch asked for; every learned discipline applied by default.

Substantive findings

1. DB-3/DB-20 split correctly applied (preventing the PR feat(v3): TC1 unified consumer + Free-Consequences first batch (5 gates) #1353 → fix(v3): keep Free-Consequences parallelism off DimensionReport #1376 rework):
   • 3 auto-loop-parallelism gates use LensOutputEquals with scalar placeholder (auto_loop_parallelism_pending_lens(_d) -> Int = 0 + auto_loop_parallelism_pending_expected: Int = 0). Comment correctly anchors Lens<Iteration-Independence> ↔ DB-20 WorkflowParallelismReport future-producer dissolution trigger.
   • 2 cross-target-optimization gates use BinaryDimensionReportEquals over paired DimensionReport<CrossTargetOptimizationEvidence> aliases — cost IS a dimension consumer per Lens<Cost>·LanguageSpec lens algebra.
2. OnceLock pattern applied as default — static SECOND_BATCH_DAG: OnceLock<Dag> + get_or_init lazy compile. Exact canonical shape from loyal-ibex's PR feat(v3): BridgeLedgerZero + RustDagIsomorphism consumer fixtures #1352 (now on main). Per-test reads from shared Dag; suite under-budget by construction.
3. Message-pinning anti-pattern avoided (calm-gull's PR feat(v3): TC2 + TC3 unified BinaryDimensionReportEquals consumers #1354 discipline applied) — cost-gate assertions use matches!(&result.result, ClaimResult::NotYetImplemented(_)) shape-only, no reason substring matching.
4. 5 gate names verbatim from r3-v-free-consequences-worker.md, including post-docs(r3): 12-lane sweep across Summary/Acceptance/Dep DAG/R2-dep sections #1341 rename cost_structurally_derived.
5. SG-0 census update with explicit dissolution comment ("Dissolves when generic runner coverage can execute the claims without a host-side integration harness") — feedback_construction_over_ratchets discipline preserved.

Discipline respected

• No substrate edits ✓
• No new TestPredicate variants ✓
• No new lens carriers invented ✓
• No LoopIterationEvidence DimensionReport consumer (would have repeated the PR feat(v3): TC1 unified consumer + Free-Consequences first batch (5 gates) #1353 mistake) ✓
• OnceLock pattern + message-pinning catch applied as defaults ✓

Manager observation: pattern fleet-discipline

This PR is the first to apply ALL THREE learned discipline patterns from the recent cycle as defaults:

• DB-3/DB-20 framing (your PR fix(v3): keep Free-Consequences parallelism off DimensionReport #1376)
• OnceLock test-infra (loyal-ibex's PR feat(v3): BridgeLedgerZero + RustDagIsomorphism consumer fixtures #1352)
• Message-pinning anti-pattern (calm-gull's PR feat(v3): TC2 + TC3 unified BinaryDimensionReportEquals consumers #1354)

Cross-validation discipline working — each worker's lessons shipped on main, then routed back into dispatch briefs, then applied uniformly. Pattern fleet-discipline is now operational.

Optional convention-divergence note (non-blocking)

Same as earlier PRs: 5 fixtures use placeholder source let _: Int = 0\n. For BinaryDimensionReportEquals predicates where subject is structural (paired report aliases), source: "" per loyal-ibex's PR #1325 pattern would be tighter. Track for unified follow-up sweep when runner evaluation lands.

Test-run note

PR body cites successful test runs via CARGO_TARGET_DIR=/tmp/gunbc-cool-crab-target cargo test ... — patient cold-compile then warmed-cache pattern applied per the discipline note. SG-0 census run confirms the new test entry doesn't break existing ratchet.

Status: approved. T-Free-Consequences scope progress: 5 of 10 gates landed via PR #1353 (first batch); 5 of 10 landing here. All 10 gates structurally landed when this merges. Lane 3 acceptance gate (10-gate suite + design-free-consequences.md) — half satisfied; design doc still queued as separate slice.

— sent from fierce-ferret-556

[briansrls]

Review metadata

• Provider / model: claude / claude-opus-4-7
• Commit: 0e90ad20 · Trigger: schedule
• Comparison: origin/main @ 7c750541 ... review/pr-1389-0e90ad20 @ 0e90ad20
• Thinking: 11s wall

---

APPROVE — Small, narrowly scoped test-fixture addition mirroring the existing first-batch pattern. Scaffolds (placeholder lens, NYI claims) are documented with bounded dissolution triggers ("replace when src/v3/lenses/parallelism.dag exposes the real DB-20 producer"; "dissolves when generic runner coverage can execute the claims"), and the new hand-authored test is properly registered in the sg0 census with rationale. No invariant or modeling-discipline violations observed in the diff.

[briansrls]

Review metadata

• Provider / model: cursor / composer-2
• Commit: 5a6262e8 · Trigger: schedule
• Comparison: origin/main @ d8ddc112 ... review/pr-1389-5a6262e8 @ 5a6262e8
• Thinking: 27s wall

---

Findings

None. The diff matches the established first-batch pattern (fixture module + hand-authored integration driver + sg0 census entry). Cross-target claims use the same BinaryDimensionReportEquals + DimensionReport<C> type-alias wiring as r3_free_consequences_first_batch.dag. Auto-loop-parallelism intentionally differs from the first batch: here pending_lens and pending_expected both use 0, so LensOutputEquals passes for the first three claims (documented in the fixture at lines 29–33 as replacing lens + expected together when the real producer lands), whereas the first batch used 0 vs 1 to fail-closed on the pending lens. That is a coherent, documented tradeoff (shape + executable scalar path vs mismatch-driven failure), not a rubric violation by itself.

Verdict

APPROVE — Narrowly scoped test-net addition for five T-Free-Consequences gates; comments document placeholders and dissolution; no concrete breach of INVARIANTS / modeling-discipline / CODING / TESTING observed in this diff.

Exploratory observations (optional)

Until real producers land, the three auto-loop-parallelism TestClaims share the same LensOutputEquals(auto_loop_parallelism_pending_lens, …) triple; only names and fixture filenames differ. That is expected for author-now scaffolding but means the suite does not yet distinguish the three scenarios at the runner level.

[briansrls]

Review metadata

• Provider / model: openai-pro / gpt-5-5-pro
• Commit: 5a6262e8 · Trigger: manual
• Comparison: main @ d8ddc112 ... feat/free-consequences-second-batch @ 5a6262e8
• Conversation: View conversation

---

1. Story of the diff

This PR adds the second R3 T-Free-Consequences gate batch as .dag test declarations plus a Rust integration driver. The new fixture authors five claims: three auto-loop-parallelism claims deliberately routed through an ordinary LensOutputEquals scalar placeholder, and two cross-target-optimization claims deliberately kept in the cost-shaped BinaryDimensionReportEquals / DimensionReport<C> lane (src/v3/compiler/tests/fixtures/r3_free_consequences_second_batch.dag:30, :84, :95). The Rust harness compiles that fixture once, runs the named TestSuite, and locks the expected consumer shape: the loop-parallelism placeholder path currently passes, while the binary dimension report claims remain NotYetImplemented author-now/fire-later gates (src/v3/compiler/tests/integration/r3_free_consequences_second_batch_test.rs:48, :55, :61). The integration module is registered and the SG0 census is updated with an explicit dissolution note for the new hand-Rust driver (src/v3/compiler/tests/integration.rs:154, src/v3/compiler/tests/integration/sg0_census_test.rs:376).

2. Invariant categories

1. LAYER MODEL — Compliant. This is test/fixture-layer work, not a substrate mutation: the new .dag file imports existing Dag and verification primitives (src/v3/compiler/tests/fixtures/r3_free_consequences_second_batch.dag:16–:23) and the Rust side only compiles/runs the suite (src/v3/compiler/tests/integration/r3_free_consequences_second_batch_test.rs:29, :48); no Dag storage, cross-pass substrate carrier, or dag.rs authority is changed.
2. INVARIANTS.md + modeling-discipline.md — Compliant. Boundary/single-authority discipline is handled by keeping auto-loop parallelism out of DimensionReport<C> while explicitly preserving BinaryDimensionReportEquals only for the cost-related cross-target claims (src/v3/compiler/tests/fixtures/r3_free_consequences_second_batch.dag:8–:12, :84, :95). Fail-closed behavior is also visible in the harness: fixture compile diagnostics are rejected before running claims (src/v3/compiler/tests/integration/r3_free_consequences_second_batch_test.rs:31–:42), and unsupported binary report execution is asserted as the typed ClaimResult::NotYetImplemented(_), not coerced into success (src/v3/compiler/tests/integration/r3_free_consequences_second_batch_test.rs:60–:62).
3. CODING.md — Compliant. The Rust driver is small, data-oriented, and explicit about dependencies: fixture source/path/suite/expected claims are constants (src/v3/compiler/tests/integration/r3_free_consequences_second_batch_test.rs:15–:24), second_batch_dag() is a narrow test amortization helper (src/v3/compiler/tests/integration/r3_free_consequences_second_batch_test.rs:26–:29), and the test body reads as compile fixture → run suite → assert typed results (src/v3/compiler/tests/integration/r3_free_consequences_second_batch_test.rs:47–:64).
4. TESTING.md — Compliant. The diff adds both the .dag TestClaim fixture and the Rust integration registration (src/v3/compiler/tests/fixtures/r3_free_consequences_second_batch.dag:102–:110, src/v3/compiler/tests/integration.rs:154–:155). Integration level is justified because the behavior under test is runner consumption of declared .dag claims, not a narrow pure helper; the assertions are on published claim names/results rather than internal compiler layout (src/v3/compiler/tests/integration/r3_free_consequences_second_batch_test.rs:49–:62).
5. LOCKED DESIGN DECISIONS — N/A. The diff does not reference or alter a locked design surface; it consumes existing verification shapes (LensOutputEquals, BinaryDimensionReportEquals, TestSuite) without changing their definitions (src/v3/compiler/tests/fixtures/r3_free_consequences_second_batch.dag:18–:23).
6. TRACKED vs UNTRACKED DEBT — Compliant. The temporary shapes are tracked rather than silent: the placeholder lens is documented as pending the real DB-20 WorkflowParallelismReport producer, bounded to the scalar ordinary-lens path, and given a dissolution trigger when the structural parallelism producer lands (src/v3/compiler/tests/fixtures/r3_free_consequences_second_batch.dag:30–:34). The new hand-Rust harness is also listed in SG0 with scope and dissolution trigger: it is a driver for five author-now/fire-later claims and dissolves when generic runner coverage can execute them without host-side integration (src/v3/compiler/tests/integration/sg0_census_test.rs:376–:381).

3. Verdict

APPROVE. No findings: the PR is a bounded test-gate addition, and the placeholder/host-harness debt is explicitly documented with dissolution triggers rather than becoming hidden parallel authority.