90000x · 09 — Regression Matrix

Phase: I (regression component) Status: baseline captured; SSOT bumps and pinning-test updates DEFERRED until operator applies 037–041.

Live counts (2026-05-25, captured this run)

Surface	Pre-apply (live)	Authored delta	Post-apply (expected)	SSOT site to bump
dot_iu_command_catalog rows	30	+6	36	cutter_agent/iu_core/dot_commands.py::_REGISTRY
iu_core_retention_policy rows	4	+2	6	pinning test in 540x suite
iu_core.iu_staging_record rows	3	0	3 (until proof step 1 inserts)	n/a
iu_core.iu_staging_payload rows	4	0	4 (until proof step 1 inserts 3)	n/a
CHECK constraints on iu_core.iu_staging_record	7	+1 swap, +1 swap, +1 new	8 (lifecycle swap, consumed swap, expiry-ceiling new)	n/a
Functions in public schema (sample)	fn_iu_compose, fn_iu_reconstruct_source, fn_iu_subtree, fn_iu_collection_add_piece, fn_iu_composer_enabled, fn_iu_core_retention_cleanup, fn_iu_staging_healthcheck	+6	+6	D9 expected_fn 67 → 73 (assumed 67 baseline; verify pre-apply)
DOT-tier objects in D9	(TBD via D9 healthcheck)	+6	+6	D9 expected_dot

Note: this macro intentionally did NOT bump SSOT (_REGISTRY) or pinning tests, per [[feedback-honest-channel-block-beats-partial-trigger]] and [[feedback-pinning-tests-bump-per-macro]] — the rule is "bump SSOT in the same commit as apply, not at authoring time".

DIEU axis invariants (must be preserved post-apply)

Axis	Expected count	Source
DIEU-28	27	70000x memory baseline
DIEU-32	23	70000x memory baseline
DIEU-35	36	70000x memory baseline

fn_iu_verify_cut does NOT check these directly (it's per-run), so an operator-side post-proof query is part of the regression block:

SELECT dieu_code, count(*) FROM iu.information_unit
WHERE dieu_code IN ('DIEU-28','DIEU-32','DIEU-35') GROUP BY dieu_code ORDER BY dieu_code;

If counts change after 037-041 apply (no proof yet), STOP — apply caused unintended axis drift.

D9 conformance expected state

Pre-apply baseline (from 70000x memory + 50000x adjustment): 181 object D9 surface, 7/7 PASS. Each new fn/DOT bumps expected_* counters by 1. After 037-041 + SSOT bumps:

• 6 new fns → fn count bump
• 6 new DOTs → DOT count bump
• 2 new retention rows → retention count bump
• 3 CHECK constraint changes → not D9-counted (CHECKs are not catalog-tier)

Operator commit message should read:

apply migrations 037-041 + 90000x mark→cut pipeline; SSOT bumps 30→36 DOT, 67→73 fn, 4→6 retention; D9 expected 181→[new].

Pinning-test bumps required (post-apply)

Pinning test	Site	Old	New
test_dot_command_catalog_size	tests/pinning/test_dot_command_catalog.py	30	36
test_retention_policy_count	tests/pinning/test_retention.py	4	6
test_staging_lifecycle_vocab	NEW — tests/pinning/test_staging_lifecycle.py	n/a	check {pending,pending_review,approved,consumed,rejected,expired,cleaned}
test_no_vector_excluded_invariant	tests/pinning/test_no_vector.py	exists	extend to assert 4 covering surfaces (CHECK + sidecar 190/191 + view + iu_vector_sync_point shape)
test_d9_expected_fn_count	D9 conformance suite	67 (per memory)	73
test_d9_expected_dot_count	D9 conformance suite	(TBD)	+6

What stays UNCHANGED (must verify post-apply)

Object	Pre-count	Post-expected	Verification SQL
iu.information_unit total	(per 70000x: 175)	unchanged (unless proof step 4 ran live, then +2 from fixture)	SELECT count(*) FROM iu.information_unit;
directus.production_documents	(per 25000x: 9456+drift)	unchanged	SELECT count(*) FROM production_documents;
iu_vector_sync_point	152 (live)	unchanged	SELECT count(*) FROM iu_vector_sync_point;
Qdrant iu_core_iu_chunks	149 (per memory)	unchanged	external (Qdrant API)
entity_species SPE-NVS	id=51 present	present	SELECT id FROM entity_species WHERE entity_code='SPE-NVS';
collection_registry COL-IUS-001/002	ids 190/191	unchanged	live id select
birth_registry SPE-NVS/COL-IUS-001/002	3 rows, all status=born	unchanged	live select

Tests not yet runnable

The macro is intended to bump tests 1305 → 1305+ and D9 7/7 → 7/7 [new total]. Since 037-041 are authored-only, tests stay at their current passing level. The operator must:

1. apply 037-041,
2. update _REGISTRY and pinning literals,
3. run pytest tests/ (expecting 1305 + new test count),
4. run D9 healthcheck (SELECT fn_iu_healthcheck_d9() or equivalent),
5. then commit.

Healthcheck

The existing fn_iu_staging_healthcheck should be called post-apply:

SELECT * FROM fn_iu_staging_healthcheck();

If this fn schema is unknown, a Carry-Forward task is to author its expected output for 90000x post-state.

Cross-links

• [[feedback-pinning-tests-bump-per-macro]] — bump rule.
• [[feedback-honest-channel-block-beats-partial-trigger]] — why SSOT stays pinned at live D9 surface.
• 70000x reports — D9 conformance baseline.