Tier 18.4: mapping CSV ROW_MISMATCH polish (post-audit followup #4 FINAL)

[jamezycesar-collab]

Milestone

This PR closes the last item on the post-audit followup list. With this merged, all four post-Tier-17.5 followups are complete and the cat-data-model is in a fully steady-state.

#	Item	PR
1	Validator regex hardening	#24
2	Empty allowlist housekeeping + README	#25
3	Multi-leg Fabric Lakehouse Gold variant	#26
4	Mapping CSV ROW_MISMATCH polish	this PR

Summary

Corrects 12 row-number references on fact_multileg_option_events rows in ddl/gold/06_cat_field_mapping.csv where the PDF-verified row number differs from the row originally claimed. All 12 corrections were identified during Tier 17.0's PDF verification pass and explicitly deferred to optional polish per Tier 17.5's CHANGELOG.

Files changed

• ddl/gold/06_cat_field_mapping.csv - 12 row-number corrections
• CHANGELOG.md - Tier 18.4 entry + steady-state summary

Corrections in §5.2.2 (Multi-Leg Order Route Event)

Column	Old → New
sender_imid	row 14 → 13
destination	row 15 → 14
destination_type	row 16 → 15
routed_order_id	row 17 → 16
session	row 18 → 17
route_rejected_flag	row 27 → 26
exch_origin_code	row 28 → 27
paired_order_id	row 34 → 28

Corrections in §5.2.6 / §5.2.7

Column	Old → New
prior_order_id	§5.2.6 row 10 → 9
initiator	§5.2.6 row 22 → 21
leaves_qty	§5.2.6 row 27 → 25
cancel_qty	§5.2.7 row 13 → 12

What's NOT in scope

10 SECTION_MISMATCH cases identified in Tier 17.0 remain intentionally unmodified. They represent legitimate alternative references between:

• Multi-leg event usage sites (§5.2.x) — where the field appears in a specific multi-leg event's field table
• Single-leg canonical definition sites (§5.1.x) — where the field is first defined for option events

Both reference styles are spec-supported. The current §5.2.x style is meaningful in the multi-leg context and was preserved as a design decision.

Verification

• All 8 guardrails pass.
• 10/10 regression tests in test__ddl_parser.py pass.
• All 12 corrected rows match PDF spot-check exactly.

Post-followup state — STEADY

F3.1 phantom-table backlog:        0 (closed Tier 13-16)
F3.2 missing-column backlog:       0 (closed Tier 17.1-17.5)
Post-audit tooling debt:           0 (closed Tier 18.1-18.4)
Dialect coverage (option):         4 (Delta, Hive, Fabric Lakehouse, Fabric Warehouse)
Dialect coverage (multi-leg):      4 (same)
Guardrails passing:              8/8 with 0 field-mapping allowlist exceptions
Field-mapping rows verified:     257

Test plan

• All 8 guardrails pass (exit 0)
• 10/10 test__ddl_parser.py regression tests pass
• All 12 corrected mapping CSV rows match PDF spot-check
• No code or DDL changes (mapping CSV is documentation/metadata)
• Recurrence-safe precheck in push script