diff --git a/ARCHITECTURE.md b/ARCHITECTURE.md
index 2f23dd6d0..9b7f53ed7 100644
--- a/ARCHITECTURE.md
+++ b/ARCHITECTURE.md
@@ -86,12 +86,8 @@ flowchart LR
 | `commitment_extraction.py` | Pluggable LLM derivation of a customer commitment (promise + deadline) from a post; `Null` default, `ContextualOrchestrator` real impl |
 | `temporal_expressions.py` | Pure Korean relative-time resolver for Global Ask (ADR 0150) |
 | `ask_time_axis.py` | Event-time vs ingestion-time clock choice for that window (ADR 0202) |
-| `ontology.py` | Loads the governed Turtle source tree (`lineageweave-kg.ttl` plus generated fragments), the formal OWL 2/RDFS/SKOS vocabulary for the Knowledge Graph's node/edge types, source taxonomies, and published O*NET linkages (ADR 0004, ADR 0252, ADR 0255, ADR 0256) |
-| `backend/app/occupation_rating_ingestion.py` | Projects authenticated occupation-rating evidence plus persisted source and represented-occupation catalogs (ADR 0258, ADR 0260, ADR 0261) |
-| `frontend/src/components/OccupationRatingProfile.tsx` | Selects an imported source, filters stored occupation titles without ranking, and reads exact Dashboard evidence while preserving absence, uncertainty, and warning semantics (ADR 0259–0262) |
+| `ontology.py` | Loads `docs/ontology/lineageweave-kg.ttl`, the formal OWL 2/RDFS/SKOS vocabulary for the Knowledge Graph's node/edge types (ADR 0004) |
 | `ontology_neighborhood.py` | Bounded typed ontology/provenance neighborhood (ADR 0184); PostgreSQL stays authoritative, OWL subclass is not an instance edge |
-| `occupational_construct_catalog.py` | Official O*NET 31.0 construct catalog sync (ADR 0250); no ratings or invented IRIs |
-| `backend/app/occupational_construct_search.py` | Authorized catalog-label search over assertion-backed constructs (ADR 0257); hidden Posts never mint a hit |
 | `ontology_source_cursor.py` | Opaque HMAC source-window continuation (ADR 0124); keyset pagination, never OFFSET |
 | `period_report.py` | Fit GRM/GPCM on persisted IRT rows, FIPC-select, EAP-score a period (ADR 0003 slice 3; Bock & Mislevy, 1982) |
 | `fixtures.py` | Synthetic demo dataset -- no real data ships in this repo |
@@ -379,6 +375,11 @@ governed atomic Voice with an explicit truth state and an ABAC-visible evidence
 Post. The server creates the normalized PROV-O derivation and assignment in one
 transaction; clients never submit an internal assertion id, and this route
 cannot replace the imported primary Voice.
+Imported-primary changes are retained in `source_post_voice` as non-overlapping
+half-open effective intervals (ADR 0252). The database closes the current row
+and opens the new observed primary at one statement instant; live reads select
+the open interval, cutoff reads select the containing interval, and an ontology
+continuation without an explicit cutoff uses its frozen snapshot instant.
 The bounded ontology response carries a visible Voice assignment's evidence
 Post id alongside its exact-value row. The exact-value table therefore offers
 separate carrying-Post and derivation-evidence actions; hidden evidence removes
@@ -683,8 +684,7 @@ vocabulary (`node_type`, `edge_type`, `entity_relationship_type`,
 `person_side`, `corporate_entity_level`) actually matches what the
 Ontology/Semantic-Layer claim implies.
 
-`docs/ontology/lineageweave-kg.ttl` and its deterministic governed fragments
-are a real OWL 2 / RDFS / SKOS
+`docs/ontology/lineageweave-kg.ttl` is a real OWL 2 / RDFS / SKOS
 ontology in Turtle syntax: classes for `Post`/`Person`/`CorporateEntity`
 (with `OurSidePerson`/`CounterpartyPerson` subclasses), object
 properties for each `edge_type_code` and `entity_relationship_type`
@@ -698,7 +698,7 @@ specification over it, in the same sense W3C's own stack uses "semantic
 layer" (RDFS/OWL as the governed conceptual layer over raw data), not a
 separate BI-metrics product and not a parallel triple store.
 
-`lineageweave/ontology.py` parses the Turtle source tree once with `rdflib`
+`lineageweave/ontology.py` parses the Turtle file once with `rdflib`
 (pure Python, no Rust toolchain, unlike `fast-mlsirm`) and exposes the
 vocabulary as importable IRI constants, so application code has one
 canonical name per class/property instead of re-typing lookup codes as
@@ -715,19 +715,6 @@ enforcement mechanism: a future PR that adds a new `edge_type` or
 `entity_relationship_type` code without updating the ontology fails
 this test, not just a docstring's word.
 
-### Authorized job architecture snapshots
-
-The public SOC/O*NET vocabulary and an employer's job architecture remain
-different graphs. ADR 0263 adds an organization-scoped PostgreSQL source
-boundary for private job-family/job-series snapshots: immutable source
-metadata owns normalized nodes, source-declared broader/narrower edges, and
-optional explicit bindings to a versioned external occupation scheme. An edge
-table preserves multiple-family membership; the importer rejects cycles and
-never derives a parent or binding from a label or code pattern. The snapshot
-is source evidence only. It does not create a person, post, organizational
-unit, competency, score, weight, or ontology assertion, and runtime rows never
-enter repository artifacts.
-
 ## Phase 6c: post content normalization before any LLM/embedding call
 
 The brief's latest revision calls out, explicitly, that a post body mixing
diff --git a/CHANGELOG.md b/CHANGELOG.md
index 6e57455bb..7953f47b7 100644
--- a/CHANGELOG.md
+++ b/CHANGELOG.md
@@ -9,6 +9,12 @@ All notable changes to this project are documented here. Format follows
 ### Added
 
 - Normalized Voice-of-X composition persistence (ADR 0256): every imported
+- Imported primary Voice changes now retain non-overlapping effective intervals
+  (ADR 0252), including recurring A → B → A values. Live, knowledge-cutoff,
+  and ontology-snapshot reads select the same period contract; PostgreSQL
+  rejects overlapping primary intervals instead of relying on application
+  ordering.
+- Normalized Voice-of-X composition persistence (ADR 0251): every imported
   primary voice is mirrored into `source_post_voice`; each additional voice
   requires its own PROV-O assertion and truth status. Compound lookup codes,
   keyword inference, confidence thresholds, and invented weights remain out of
diff --git a/backend/app/ontology_neighborhood_ingestion.py b/backend/app/ontology_neighborhood_ingestion.py
index f64987ec4..f008ed3fd 100644
--- a/backend/app/ontology_neighborhood_ingestion.py
+++ b/backend/app/ontology_neighborhood_ingestion.py
@@ -866,6 +866,10 @@ def neighborhood_to_payload(neighborhood: OntologyNeighborhood) -> dict[str, Any
                 "is_primary": assignment.is_primary,
                 "truth_status_code": assignment.truth_status_code,
                 "recorded_at": assignment.recorded_at.isoformat(),
+                "effective_from": assignment.effective_from.isoformat(),
+                "effective_to": assignment.effective_to.isoformat()
+                if assignment.effective_to
+                else None,
                 "provenance_reference": assignment.provenance_reference,
                 "evidence_post_id": assignment.evidence_post_id,
             }
@@ -890,6 +894,7 @@ async def _load_voice_assignments(
         """
         select voice.post_id, voice.voice_type_code, lookup.lookup_label, voice.is_primary,
                voice.truth_status_code, voice.recorded_at,
+               voice.effective_from, voice.effective_to,
                case when evidence.node_id = any($1::uuid[]) then evidence.node_id end
                    as evidence_post_id
           from source_post_voice voice
@@ -903,10 +908,11 @@ async def _load_voice_assignments(
            and evidence.node_type_code = 'node_post'
          where voice.post_id = any($1::uuid[])
            and (voice.is_primary or evidence.node_id = any($1::uuid[]))
-           and (($2::timestamptz is null and voice.effective_to is null)
-                or ($2::timestamptz is not null
-                    and voice.effective_from <= $2
-                    and (voice.effective_to is null or $2 < voice.effective_to)))
+           and voice.effective_from <= coalesce($2::timestamptz, $3::timestamptz)
+           and (
+               voice.effective_to is null
+               or coalesce($2::timestamptz, $3::timestamptz) < voice.effective_to
+           )
            and voice.recorded_at <= $3::timestamptz
          order by voice.post_id, voice.is_primary desc,
                   lookup.display_order, voice.voice_type_code
@@ -931,6 +937,8 @@ async def _load_voice_assignments(
                 is_primary=row["is_primary"],
                 truth_status_code=row["truth_status_code"],
                 recorded_at=row["recorded_at"],
+                effective_from=row["effective_from"],
+                effective_to=row["effective_to"],
                 provenance_reference=(
                     "Evidence-backed additional voice"
                     if not row["is_primary"]
diff --git a/backend/tests/test_api.py b/backend/tests/test_api.py
index 4cb798a69..7500bfcc0 100644
--- a/backend/tests/test_api.py
+++ b/backend/tests/test_api.py
@@ -15,10 +15,11 @@
 import asyncio
 import math
 import os
-import subprocess
 import uuid
+from concurrent.futures import ThreadPoolExecutor
 from contextlib import closing
 from pathlib import Path
+from threading import Barrier
 from types import SimpleNamespace
 
 import asyncpg
@@ -38,6 +39,11 @@
 _VALKEY_URL = os.environ.get("LINEAGEWEAVE_TEST_VALKEY_URL", "redis://localhost:16379/0")
 _REALM = "lineageweave-demo"
 _MIGRATION_PATH = Path(__file__).resolve().parents[2] / "migrations" / "0001_initial_schema.sql"
+_PROVENANCE_MIGRATION = (
+    Path(__file__).resolve().parents[2]
+    / "migrations"
+    / "0017_prov_o_standard_relations.sql"
+)
 _REGISTRY_MIGRATION = Path(__file__).resolve().parents[2] / "migrations" / "0018_analysis_run_registry.sql"
 _RETENTION_MIGRATION = Path(__file__).resolve().parents[2] / "migrations" / "0020_analysis_run_retention_purge.sql"
 _RECONSTRUCTION_MIGRATION = (
@@ -88,6 +94,9 @@
 _SOURCE_ORG_NAMED_HINTS_MIGRATION = (
     Path(__file__).resolve().parents[2] / "migrations" / "0039_source_org_named_hints.sql"
 )
+_VOC_VOCABULARY_MIGRATION = (
+    Path(__file__).resolve().parents[2] / "migrations" / "0042_voc_type_vocabulary.sql"
+)
 _MEMBER_LOCALE_MIGRATION = (
     Path(__file__).resolve().parents[2] / "migrations" / "0044_member_locale_preference.sql"
 )
@@ -182,6 +191,11 @@
     / "migrations"
     / "0206_report_leftover_map_reconstruction.sql"
 )
+_LEFTOVER_MAP_UNEXPLAINED_SHARE_MIGRATION = (
+    Path(__file__).resolve().parents[2]
+    / "migrations"
+    / "0233_report_leftover_map_unexplained_share.sql"
+)
 _GLOBAL_ASK_JOB_MIGRATION = (
     Path(__file__).resolve().parents[2]
     / "migrations"
@@ -232,6 +246,36 @@
     / "migrations"
     / "0183_source_post_event_occurred_at.sql"
 )
+_ONTOLOGY_TRUTH_STATUS_MIGRATION = (
+    Path(__file__).resolve().parents[2]
+    / "migrations"
+    / "0175_ontology_truth_status.sql"
+)
+_VOICE_TAXONOMY_MIGRATION = (
+    Path(__file__).resolve().parents[2]
+    / "migrations"
+    / "0235_voice_of_x_post_taxonomy.sql"
+)
+_VOICE_ASSIGNMENT_MIGRATION = (
+    Path(__file__).resolve().parents[2]
+    / "migrations"
+    / "0237_source_post_voice_combination.sql"
+)
+_VOICE_HISTORY_MIGRATION = (
+    Path(__file__).resolve().parents[2]
+    / "migrations"
+    / "0243_source_post_voice_history.sql"
+)
+_OCCUPATIONAL_CONSTRUCT_MIGRATION = (
+    Path(__file__).resolve().parents[2]
+    / "migrations"
+    / "0238_occupational_construct_assertion.sql"
+)
+_OCCUPATIONAL_CATALOG_MIGRATION = (
+    Path(__file__).resolve().parents[2]
+    / "migrations"
+    / "0239_occupational_construct_catalog.sql"
+)
 
 
 def _postgres_available() -> bool:
@@ -317,6 +361,7 @@ def seeded_db(demo_analyst_token):
     try:
         with conn.cursor() as cur:
             cur.execute(_MIGRATION_PATH.read_text())
+            cur.execute(_PROVENANCE_MIGRATION.read_text())
             cur.execute(_REGISTRY_MIGRATION.read_text())
             cur.execute(_RETENTION_MIGRATION.read_text())
             cur.execute(_RECONSTRUCTION_MIGRATION.read_text())
@@ -339,6 +384,7 @@ def seeded_db(demo_analyst_token):
                 (Path(__file__).resolve().parents[2] / "migrations" / "0040_post_summary_contract.sql")
                 .read_text()
             )
+            cur.execute(_VOC_VOCABULARY_MIGRATION.read_text())
             cur.execute(_MEMBER_LOCALE_MIGRATION.read_text())
             cur.execute(_IMAGE_REGION_MIGRATION.read_text())
             cur.execute(_POST_CONTENT_STRUCTURE_MIGRATION.read_text())
@@ -402,11 +448,18 @@ def seeded_db(demo_analyst_token):
             cur.execute(_GLOBAL_ASK_KNOWLEDGE_CUTOFF_MIGRATION.read_text())
             cur.execute(_GLOBAL_ASK_PUBLIC_VERIFICATION_MIGRATION.read_text())
             cur.execute(_EVENT_OCCURRED_AT_MIGRATION.read_text())
+            cur.execute(_ONTOLOGY_TRUTH_STATUS_MIGRATION.read_text())
             cur.execute(_LEFTOVER_MAP_AXIS_MIGRATION.read_text())
             cur.execute(_CHANNEL_EVIDENCE_MIGRATION.read_text())
+            cur.execute(_OCCUPATIONAL_CONSTRUCT_MIGRATION.read_text())
+            cur.execute(_OCCUPATIONAL_CATALOG_MIGRATION.read_text())
             cur.execute(_LEFTOVER_MAP_UNEXPLAINED_MIGRATION.read_text())
             cur.execute(_LEFTOVER_MAP_CROSS_SHARE_MIGRATION.read_text())
             cur.execute(_LEFTOVER_MAP_RECONSTRUCTION_MIGRATION.read_text())
+            cur.execute(_VOICE_TAXONOMY_MIGRATION.read_text())
+            cur.execute(_VOICE_ASSIGNMENT_MIGRATION.read_text())
+            cur.execute(_VOICE_HISTORY_MIGRATION.read_text())
+            cur.execute(_LEFTOVER_MAP_UNEXPLAINED_SHARE_MIGRATION.read_text())
             cur.execute(
                 "insert into common_lookup_value (lookup_category, lookup_code, lookup_label) values "
                 "('corporate_entity_level', 'group', 'Group'), "
@@ -414,7 +467,6 @@ def seeded_db(demo_analyst_token):
                 "('corporate_entity_level', 'plant', 'Plant'), "
                 "('post_visibility', 'public', 'Public'), "
                 "('post_visibility', 'private', 'Private'), "
-                "('voc_type', 'voc', 'Voice of Customer'), "
                 "('permission', 'post_read', 'Read posts'), "
                 "('person_side', 'our_side', 'Our side'), "
                 "('person_side', 'counterparty', 'Counterparty'), "
@@ -1864,6 +1916,20 @@ def test_post_list_includes_public_and_own_corp_but_excludes_other_corp(client,
     assert public["voc_type_label"] == "Voice of Customer"
     assert public["visibility_label"] == "Public"
     assert {option["code"] for option in payload["voc_type_options"]} == {"voc"}
+    assert {option["code"] for option in payload["voice_type_catalog"]} == {
+        "voc",
+        "voe",
+        "vob",
+        "voi",
+        "vop",
+        "vops",
+        "vos",
+        "vocc",
+        "voco",
+        "vom",
+        "vor",
+        "voso",
+    }
     assert {option["code"] for option in payload["visibility_options"]} == {"public", "private"}
     assert next(option for option in payload["visibility_options"] if option["code"] == "public")["label"] == "Public"
 
@@ -1960,6 +2026,100 @@ def test_post_detail_exposes_explicit_and_semantic_project_evidence(
     assert listed_post["project_evidence"][0]["provenance"] == "post_project_mention.evidence_text"
 
 
+def test_post_detail_exposes_evidence_bound_occupational_construct(
+    client, demo_analyst_token, seeded_db
+) -> None:
+    """The authorized detail projection preserves its exact synthetic evidence."""
+    conn = psycopg2.connect(seeded_db["dsn"])
+    try:
+        with conn.cursor() as cur:
+            cur.execute(
+                """
+                insert into post_content_unit
+                    (post_id, unit_index, unit_kind_code, unit_text)
+                values (%s, 90, 'plain_text', %s)
+                returning post_content_unit_id
+                """,
+                (
+                    seeded_db["public_post_id"],
+                    "Synthetic record requires oral comprehension.",
+                ),
+            )
+            unit_id = cur.fetchone()[0]
+            cur.execute(
+                """
+                insert into occupational_construct_vocabulary
+                    (vocabulary_iri, version_label, license_iri, attribution_text)
+                values (%s, '31.0', %s, 'Synthetic O*NET attribution')
+                returning vocabulary_id
+                """,
+                (
+                    "https://www.onetcenter.org/database.html",
+                    "https://creativecommons.org/licenses/by/4.0/",
+                ),
+            )
+            vocabulary_id = cur.fetchone()[0]
+            cur.execute(
+                """
+                insert into occupational_construct
+                    (vocabulary_id, construct_iri, construct_family_code, preferred_label)
+                values (%s, %s, 'cognitive_ability', 'Oral Comprehension')
+                returning construct_id
+                """,
+                (vocabulary_id, "https://data.onetcenter.org/element/1.A.1.a.1"),
+            )
+            construct_id = cur.fetchone()[0]
+            cur.execute(
+                """
+                insert into post_occupational_construct_assertion
+                    (post_id, post_content_unit_id, construct_id, evidence_text,
+                     truth_status_code, extraction_method, orchestrator_session_id)
+                values (%s, %s, %s, 'oral comprehension', 'truth_inferred',
+                        'contextual_orchestrator_structured', 'synthetic-session')
+                """,
+                (seeded_db["public_post_id"], unit_id, construct_id),
+            )
+        conn.commit()
+    finally:
+        conn.close()
+
+    response = client.get(
+        f"/api/posts/{seeded_db['public_post_id']}",
+        headers={"Authorization": f"Bearer {demo_analyst_token}"},
+    )
+    assert response.status_code == 200
+    assertions = response.json()["occupational_construct_assertions"]
+    assert assertions[0]["preferred_label"] == "Oral Comprehension"
+    assert assertions[0]["evidence_text"] == "oral comprehension"
+    assert assertions[0]["provenance"] == (
+        "post_occupational_construct_assertion.evidence_text"
+    )
+
+    conn = psycopg2.connect(seeded_db["dsn"])
+    try:
+        with conn.cursor() as cur:
+            cur.execute(
+                """
+                update post_content_ingestion_job
+                   set source_body_sha256 = %s,
+                       status_code = 'post_content_ingestion_failed'
+                 where post_id = %s
+                """,
+                ("f" * 64, seeded_db["public_post_id"]),
+            )
+        conn.commit()
+    finally:
+        conn.close()
+
+    stale = client.get(
+        f"/api/posts/{seeded_db['public_post_id']}",
+        headers={"Authorization": f"Bearer {demo_analyst_token}"},
+    )
+    assert stale.status_code == 200
+    assert stale.json()["occupational_construct_evidence_status"] == "unavailable"
+    assert stale.json()["occupational_construct_assertions"] == []
+
+
 def test_post_detail_as_of_returns_the_cutoff_known_body(
     client, demo_analyst_token, seeded_db
 ) -> None:
@@ -4658,6 +4818,165 @@ def _grant_post_admin(dsn: str) -> None:
         admin_conn.close()
 
 
+def test_create_voice_assignment_persists_authorized_prov_o_evidence(
+    client, demo_analyst_token, seeded_db
+) -> None:
+    """A real OIDC admin write retains its governed Voice and evidence Post."""
+    headers = {"Authorization": f"Bearer {demo_analyst_token}"}
+    endpoint = f"/api/posts/{seeded_db['own_private_post_id']}/voice-assignments"
+    payload = {
+        "voice_type_code": "vos",
+        "truth_status_code": "truth_observed",
+        "evidence_post_id": seeded_db["public_post_id"],
+    }
+
+    assert client.post(endpoint, json=payload, headers=headers).status_code == 403
+    _grant_post_admin(seeded_db["dsn"])
+    with closing(psycopg2.connect(seeded_db["dsn"])) as conn, conn.cursor() as cur:
+        cur.execute(
+            "insert into provenance_resource (resource_iri, resource_label) "
+            "values (%s, 'Synthetic evidence Post') returning resource_id",
+            (f"https://example.test/posts/{seeded_db['public_post_id']}",),
+        )
+        resource_id = cur.fetchone()[0]
+        cur.execute(
+            "insert into provenance_resource_type (resource_id, class_code) "
+            "values (%s, 'prov_entity')",
+            (resource_id,),
+        )
+        cur.execute(
+            "insert into provenance_resource_binding "
+            "(resource_id, node_type_code, node_id) values (%s, 'node_post', %s)",
+            (resource_id, seeded_db["public_post_id"]),
+        )
+        conn.commit()
+
+    response = client.post(endpoint, json=payload, headers=headers)
+    assert response.status_code == 201, response.text
+    assert response.json() == {
+        "code": "vos",
+        "label": "Voice of Supplier",
+        "is_primary": False,
+        "truth_status_code": "truth_observed",
+        "evidence_available": True,
+    }
+
+    with closing(psycopg2.connect(seeded_db["dsn"])) as conn, conn.cursor() as cur:
+        cur.execute(
+            """
+            select voice.voice_type_code, voice.is_primary,
+                   assertion.relation_code, binding.node_id
+              from source_post_voice voice
+              join provenance_assertion assertion
+                on assertion.assertion_id = voice.provenance_assertion_id
+              join provenance_resource_binding binding
+                on binding.resource_id = assertion.object_resource_id
+             where voice.post_id = %s and voice.voice_type_code = 'vos'
+            """,
+            (seeded_db["own_private_post_id"],),
+        )
+        stored = cur.fetchone()
+        assert stored == (
+            "vos",
+            False,
+            "prov_was_derived_from",
+            seeded_db["public_post_id"],
+        )
+        cur.execute(
+            "select voice_type_code from source_post_voice "
+            "where post_id = %s and is_primary",
+            (seeded_db["own_private_post_id"],),
+        )
+        assert cur.fetchone() == ("voc",)
+
+
+def test_primary_voice_history_survives_concurrent_changes_and_cutoff_reads(
+    client, demo_analyst_token, seeded_db
+) -> None:
+    """Concurrent A→B→C→A changes retain one non-overlapping primary timeline."""
+    post_id = seeded_db["own_private_post_id"]
+    barrier = Barrier(2)
+
+    with closing(psycopg2.connect(seeded_db["dsn"])) as conn, conn.cursor() as cur:
+        cur.execute(_VOICE_ASSIGNMENT_MIGRATION.read_text())
+        cur.execute(_VOICE_ASSIGNMENT_MIGRATION.read_text())
+
+    def update_voice(code: str) -> None:
+        with closing(psycopg2.connect(seeded_db["dsn"])) as conn, conn.cursor() as cur:
+            barrier.wait()
+            cur.execute(
+                "update source_post set voc_type_code = %s where post_id = %s",
+                (code, post_id),
+            )
+            conn.commit()
+
+    with ThreadPoolExecutor(max_workers=2) as executor:
+        futures = [executor.submit(update_voice, code) for code in ("vop", "voe")]
+        for future in futures:
+            future.result()
+
+    with closing(psycopg2.connect(seeded_db["dsn"])) as conn, conn.cursor() as cur:
+        cur.execute(
+            "update source_post set voc_type_code = 'voc' where post_id = %s",
+            (post_id,),
+        )
+        conn.commit()
+        cur.execute(
+            "select voice_type_code, effective_from, effective_to "
+            "from source_post_voice where post_id = %s and is_primary "
+            "order by effective_from, voice_assignment_id",
+            (post_id,),
+        )
+        history = cur.fetchall()
+
+    assert len(history) == 4
+    assert history[0][0] == "voc"
+    assert history[-1][0] == "voc"
+    assert sum(effective_to is None for _, _, effective_to in history) == 1
+    assert all(
+        history[index][2] == history[index + 1][1]
+        for index in range(len(history) - 1)
+    )
+    for voice_type_code, effective_from, _effective_to in history:
+        response = client.get(
+            f"/api/posts/{post_id}",
+            params={"as_of": effective_from.isoformat()},
+            headers={"Authorization": f"Bearer {demo_analyst_token}"},
+        )
+        assert response.status_code == 200, response.text
+        primaries = [
+            voice for voice in response.json()["voice_types"] if voice["is_primary"]
+        ]
+        assert [voice["code"] for voice in primaries] == [voice_type_code]
+
+
+def test_voice_migration_family_replays_after_a_primary_change(seeded_db) -> None:
+    """Every-start replay remains valid after a post accumulates Voice history."""
+    post_id = seeded_db["own_private_post_id"]
+    with closing(psycopg2.connect(seeded_db["dsn"])) as conn, conn.cursor() as cur:
+        cur.execute(
+            "update source_post set voc_type_code = 'vob' where post_id = %s",
+            (post_id,),
+        )
+        conn.commit()
+
+        for _ in range(2):
+            cur.execute(_VOICE_ASSIGNMENT_MIGRATION.read_text())
+            cur.execute(_VOICE_HISTORY_MIGRATION.read_text())
+            conn.commit()
+
+        cur.execute(
+            "select count(*) filter (where effective_to is null), "
+            "count(*) from source_post_voice "
+            "where post_id = %s and is_primary",
+            (post_id,),
+        )
+        current_count, history_count = cur.fetchone()
+
+    assert current_count == 1
+    assert history_count >= 2
+
+
 def test_tickets_list_is_empty_before_any_created(client, demo_analyst_token, seeded_db) -> None:
     response = client.get(
         f"/api/posts/{seeded_db['own_private_post_id']}/tickets",
@@ -5715,10 +6034,15 @@ def test_seed_period_report_surfaces_on_get_reports(client, demo_analyst_token,
         if share is not None:
             assert not math.isnan(share)
             assert not math.isinf(share)
+        unexplained_share = pair.get("leftover_map_unexplained_share")
+        assert unexplained_share is None or isinstance(unexplained_share, (int, float))
+        if unexplained_share is not None:
+            assert not math.isnan(unexplained_share)
+            assert not math.isinf(unexplained_share)
+            assert unexplained_share >= 0.0
         if unexplained is not None and reconstruction is not None:
             assert unexplained + reconstruction == pytest.approx(pair["leftover_residual"])
         assert "leftover_map_explained_share" not in pair
-        assert "leftover_map_unexplained_share" not in pair
     leftover_axes = high_report.get("leftover_map_axes", [])
     assert [axis["axis_index"] for axis in leftover_axes] == [1, 2]
     assert all(axis["leftover_singular_value"] >= 0 for axis in leftover_axes)
@@ -5781,6 +6105,11 @@ def test_seed_period_report_surfaces_on_get_reports(client, demo_analyst_token,
         or isinstance(pair["leftover_map_reconstruction"], (int, float))
         for pair in leftover_thread.get("leftover_pairs", [])
     )
+    assert all(
+        "leftover_map_unexplained_share" not in pair
+        and "leftover_map_explained_share" not in pair
+        for pair in leftover_thread.get("leftover_pairs", [])
+    )
 
 
 def test_seed_period_report_includes_fixture_event_lineage_posts(
diff --git a/docs/adr/0245-io-occupational-taxonomy-in-the-published-ontology.md b/docs/adr/0245-io-occupational-taxonomy-in-the-published-ontology.md
index a8476f33d..e556ba8ae 100644
--- a/docs/adr/0245-io-occupational-taxonomy-in-the-published-ontology.md
+++ b/docs/adr/0245-io-occupational-taxonomy-in-the-published-ontology.md
@@ -3,10 +3,7 @@
 **Status:** Accepted
 **Date:** 2026-08-26
 **Extends:** [ADR 0004](0004-knowledge-graph-ontology.md), [ADR 0145](0145-psychometric-channel-weight-estimation.md), [ADR 0207](0207-repository-case-ontology-namespace-canonical.md), [ADR 0232](0232-worker-function-taxonomy-in-the-published-ontology.md)
-<<<<<<< HEAD
-=======
 **Superseded in part by:** [ADR 0252](0252-complete-2018-soc-hierarchy.md), which expands the major-group-only scheme into the complete 2018 SOC hierarchy.
->>>>>>> origin/feat/onet-rating-occupation-filter
 
 ## Context
 
diff --git a/docs/adr/0252-temporal-primary-voice-history.md b/docs/adr/0252-temporal-primary-voice-history.md
new file mode 100644
index 000000000..2cf3f5685
--- /dev/null
+++ b/docs/adr/0252-temporal-primary-voice-history.md
@@ -0,0 +1,97 @@
+# ADR 0252: Temporal history for imported primary Voice
+
+## Status
+
+Accepted (2026-08-27). Extends ADR 0251 and closes issue #748.
+
+## Context
+
+ADR 0251 records when a Voice assignment starts, but migration 0237 deletes
+the former imported primary when `source_post.voc_type_code` changes. The live
+value is honest, yet an authorized knowledge-cutoff read after that update can
+no longer recover the primary that was effective at the cutoff. The existing
+`(post_id, voice_type_code)` key also cannot represent A → B → A.
+
+OWL-Time distinguishes instants from intervals and gives an interval explicit
+beginning and end bounds. PostgreSQL range types and exclusion constraints are
+the native database mechanism for rejecting overlapping periods. Neither
+source supplies a missing business-effective instant, so LineageWeave must not
+invent one: an imported change becomes effective at the database transaction
+instant when no source change instant exists.
+
+## Decision
+
+- Keep `source_post_voice` as the normalized assignment relation. Add nullable
+  `effective_to`; each row is a half-open interval
+  `[effective_from, effective_to)`. Null means current.
+- Change the key to `(post_id, voice_type_code, effective_from)`, allowing the
+  same atomic Voice to recur in non-overlapping periods.
+- Use PostgreSQL GiST exclusion constraints to reject overlapping primary
+  intervals for one Post. A partial unique index also permits at most one
+  current row for a `(post_id, voice_type_code)` pair.
+- When the imported primary changes, one trigger transaction closes both the
+  current primary and any current additional assignment for the incoming
+  Voice, then inserts the new observed primary at one trigger-execution
+  timestamp. PostgreSQL `clock_timestamp()` is read after the source-row lock
+  is acquired, so a waiting concurrent update cannot backdate its interval to
+  the earlier statement start. It never overwrites or fabricates the former
+  interval.
+- Live reads select `effective_to is null`. Cutoff reads select the row whose
+  interval contains the cutoff. Ontology continuation reads use their frozen
+  `snapshot_at` when no knowledge cutoff was requested, so a page minted
+  before a change cannot silently switch to the new primary.
+- Existing rows migrate as open intervals. Migration replay changes neither
+  their starts nor their history. History before ADR 0252 remains unavailable
+  because the deleted facts cannot be reconstructed honestly.
+- This is valid-time history for a source assignment, not psychometric or
+  mathematical modeling. No weight, confidence, inference, or new Voice code
+  is introduced.
+
+## Data model
+
+```mermaid
+classDiagram
+  class SourcePost {
+    uuid post_id
+    text voc_type_code
+  }
+  class SourcePostVoice {
+    uuid post_id
+    text voice_type_code
+    boolean is_primary
+    timestamptz effective_from
+    timestamptz effective_to
+    timestamptz recorded_at
+  }
+  SourcePost "1" --> "1..*" SourcePostVoice
+```
+
+```mermaid
+sequenceDiagram
+  participant Import
+  participant SourcePost
+  participant VoiceHistory
+  Import->>SourcePost: update primary A to B
+  SourcePost->>VoiceHistory: close current A after source-row lock
+  SourcePost->>VoiceHistory: close current additional B, if present
+  SourcePost->>VoiceHistory: insert observed primary B at same instant
+  VoiceHistory-->>Import: one non-overlapping current primary
+```
+
+## Consequences
+
+- A → B → A is auditable without copying source content or exposing real
+  identifiers.
+- Half-open bounds assign the exact change instant to the new primary and avoid
+  double matches.
+- The exclusion constraint adds a GiST index and write-time check. This table
+  is bounded by Voice assignments per Post; partitioning is not warranted
+  until observed volume or lock evidence shows otherwise.
+
+## References
+
+Cox, S. J. D., & Little, C. (2022). *Time ontology in OWL*. World Wide
+Web Consortium. https://www.w3.org/TR/owl-time/
+
+PostgreSQL Global Development Group. (2025). *PostgreSQL 18 documentation:
+Range types*. https://www.postgresql.org/docs/18/rangetypes.html
diff --git a/docs/adr/README.md b/docs/adr/README.md
index f10df7664..d90943d19 100644
--- a/docs/adr/README.md
+++ b/docs/adr/README.md
@@ -9,9 +9,11 @@ decision from them.
 
 | Supporting document | Normative ADR |
 |---|---|
-| [`product-requirements.md`](../product-requirements.md) | Product requirements projection across the ADR set; ADRs remain normative |
+| [`product-requirements.md`](../product-requirements.md) | Product requirements projection across the ADR set; ADRs remain normative, including [0252](0252-temporal-primary-voice-history.md) |
 | [`product-technical-gap-baseline.md`](../product-technical-gap-baseline.md) | Product/technical traceability projection across the ADR set; ADRs remain normative |
 | [`lineage-bi-research-notes.md`](../lineage-bi-research-notes.md) | [0084](0084-lineage-research-grounding.md), [0062](0062-semantic-unit-embedding.md), [0064](0064-lineage-evidence-and-tree-assembly.md), [0024](0024-rankweave-fusion-fail-closed.md), [0165](0165-quantity-script-display.md), [0167](0167-rankweave-ranking-channel-evidence.md), [0169](0169-ask-batched-lineage-graph.md), [0172](0172-event-lineage-channel-evidence.md), [0202](0202-ask-event-time-filter.md), [0223](0223-explicit-semantic-content-unit-kinds.md), [0238](0238-source-conversation-turn-import-contract.md) |
+| [`voice-combination-technical-requirements.md`](../voice-combination-technical-requirements.md) | [0246](0246-expanded-voice-of-x-post-taxonomy.md), [0251](0251-evidence-bearing-voice-combinations.md), [0252](0252-temporal-primary-voice-history.md) |
+| [`lineage-bi-research-notes.md`](../lineage-bi-research-notes.md) | [0084](0084-lineage-research-grounding.md), [0062](0062-semantic-unit-embedding.md), [0064](0064-lineage-evidence-and-tree-assembly.md), [0024](0024-rankweave-fusion-fail-closed.md), [0165](0165-quantity-script-display.md), [0167](0167-rankweave-ranking-channel-evidence.md), [0169](0169-ask-batched-lineage-graph.md), [0172](0172-event-lineage-channel-evidence.md), [0202](0202-ask-event-time-filter.md), [0223](0223-explicit-semantic-content-unit-kinds.md) |
 | [`PROV_O_IMPLEMENTATION.md`](../PROV_O_IMPLEMENTATION.md) | [0065](0065-prov-o-provenance-boundary.md) |
 | [`PROV_O_IMPLEMENTATION_MATRIX.md`](../PROV_O_IMPLEMENTATION_MATRIX.md) | [0065](0065-prov-o-provenance-boundary.md) |
 | [`ONTOLOGY_NAMESPACE_INVENTORY.md`](../doctoring/ONTOLOGY_NAMESPACE_INVENTORY.md) | [0207](0207-repository-case-ontology-namespace-canonical.md), [0157](0157-public-ontology-namespace-identity.md) |
diff --git a/docs/product-requirements.md b/docs/product-requirements.md
index a8b741521..763a116b9 100644
--- a/docs/product-requirements.md
+++ b/docs/product-requirements.md
@@ -53,6 +53,10 @@ edge exposes the same authorized endpoints and evidence through API and UI.
   governed atomic Voice only with an ABAC-visible evidence Post and explicit
   truth state; create the normalized PROV-O derivation server-side and never
   accept an opaque provenance assertion identifier from the caller.
+- Preserve recurring imported-primary history as non-overlapping half-open
+  intervals; live reads return the current primary, while authorized cutoff
+  reads return the primary effective at that instant without substituting the
+  current value.
 - Let a `post_admin` connect another perspective from the live Post popup by
   choosing an unassigned atomic Voice and an explicit truth state; use the open
   authorized Post as evidence and hide the write action on cutoff views.
@@ -63,6 +67,8 @@ edge exposes the same authorized endpoints and evidence through API and UI.
 Acceptance: Turtle, JSON-LD, N-Triples, SHACL, API payloads, persisted IRIs,
 and rendered labels agree on term kind, direction, namespace, and provenance;
 an additional Voice cannot demote the imported primary or cite hidden evidence;
+an A → B → A imported-primary sequence returns A, B, and A at its three
+respective authorized cutoff intervals;
 the exact-value table opens the carrying Post and its authorized derivation
 evidence as distinct actions;
 the authoring form has explicit selections, permission/cutoff gating, retryable
@@ -108,10 +114,8 @@ and a digest-bound run record distinguishes a supported empty result from an
 unavailable provider. ADR 0254 adds the authorized Post-detail evidence review
 surface and honest complete, processing, and unavailable states. ADR 0255
 projects assertion-backed constructs into the existing ABAC-filtered ontology
-neighborhood without duplicating graph storage or promoting truth. ADR 0257
-adds authorized catalog-label search: reviewers type an official O*NET label
-and open the earliest visible supporting Post. Constructs without visible
-evidence stay undisclosed. Occupation ratings remain unavailable.
+neighborhood without duplicating graph storage or promoting truth. Catalog
+search remains unavailable.
 
 ### PRD-FR-2C — FJA I/O-Psychology cognitive, affective & behavioral semantic layer
 
@@ -165,180 +169,6 @@ canonical namespace, and lookup round-trip isolation are enforced by
 `tests/test_io_taxonomy.py`; `tests/test_ontology.py` continues to pass
 unchanged.
 
-### PRD-FR-2A — Worker-function taxonomy
-
-- Publish the DOT/FJA Data/People/Things worker functions (24 concepts,
-  official definitions verbatim) in the canonical ontology namespace
-  (ADR 0232), each with its definitional ordinal rank. Do not infer a
-  DOT-to-O*NET or Fleishman crosswalk that the authorities do not publish.
-- Expose the taxonomy through a deterministic application read model with
-  fail-closed lookups; an absent function is an honest unknown.
-- Carry no numeric weight from the taxonomy: ranks are scale positions,
-  never calibrated weights.
-
-Acceptance: completeness, full verbatim definitions, deterministic ordering,
-and lookup round-trip isolation are enforced by
-`tests/test_worker_function_taxonomy.py`; `tests/test_ontology.py`
-continues to pass unchanged.
-
-### PRD-FR-2B — Occupational classification and worker-characteristic taxonomy
-
-- Publish all four levels of the 2018 Standard Occupational Classification:
-  23 major groups, 98 minor groups, 459 broad occupations, and 867 detailed
-  occupations with exact source parents, titles, and codes (ADR 0252), plus
-  the four O*NET 31.0 job-zone categories with
-  published names and source values 2 through 5 (ADR 0245).
-- Publish the worker-characteristic families that work-related
-  cognition, affect, and behavior resolve into: Fleishman's four ability
-  domains, Holland's six RIASEC interest types with the published
-  hexagonal adjacency relation, the six explicitly legacy O*NET work-value
-  clusters, and
-  the seven higher-order dimensions of the revised O*NET Work Styles
-  structure.
-- Publish all 3,006 O*NET 31.0 Content Model Reference elements with exact
-  identifiers, names, descriptions, and source-defined outline parents
-  (ADR 0264). Treat the six roots and 18 second-level branches as navigation
-  classes, never occupation ratings, person traits, scores, or weights.
-- Declare typed derivation properties from classifications to
-  characteristics but assert no instance binding; binding requires a
-  versioned released source profile imported with provenance in its own
-  decision.
-- Expose everything through a deterministic application read model with
-  fail-closed lookups; carry no numeric importance or level rating from
-  any occupational profile.
-
-Acceptance: completeness counts, verbatim titles, closed RIASEC
-vocabulary, exact published adjacency pairs, deterministic ordering,
-canonical namespace, and lookup round-trip isolation are enforced by
-`tests/test_io_taxonomy.py`, `tests/test_soc_2018_hierarchy.py`, and
-`tests/test_onet_content_model.py`;
-`tests/test_ontology.py` continues to pass unchanged.
-### PRD-FR-2C — Evidence-bound occupational constructs
-
-- Keep cognitive abilities, work styles, work activities, affective
-  reactions, and performance behaviors as non-equivalent construct classes
-  (ADR 0248). FJA worker functions remain separate.
-- Reuse official external identifiers and source-published relationships;
-  never infer a DPT-to-psychology crosswalk or relabel work style as affect.
-- Publish the eight O*NET 31.0 Ability, Essential Skill, Transferable Skill,
-  and Work Style link tables to Work Activities and Work Context as 1,417
-  directed, assertion-level provenance-bearing relations (ADR 0256). Treat
-  relevance as neither a causal effect nor a numeric weight.
-- Bind a construct to record content only through a provenance-bearing,
-  evidence-cited assertion. Do not promote record evidence to a person trait,
-  score, causal effect, or job requirement.
-
-Acceptance: SHACL rejects incomplete record assertions; ontology tests
-prohibit FJA equivalence, require exact Post/evidence/PROV statement structure,
-and reproduce every pinned O*NET linkage with its exact source table. Runtime
-persistence and UI remain unavailable until their separate ADR acceptance.
-
-### PRD-FR-2D — Occupation-rating source observations
-
-- Persist released occupation-to-element ratings as source observations, not
-  ontology weights: release, source table, occupation, element, scale,
-  optional category, value, sample/error/interval, suppression, relevance,
-  exact source update month, and domain source remain independently auditable
-  (ADR 0257); the product must not invent a day for O*NET's `MM/YYYY` field.
-- Keep normalized reference identities in third normal form and partition the
-  observation store by exact release then source table. An unknown partition
-  fails closed instead of entering a catch-all table.
-- Preserve decimals and missingness exactly. No local aggregation,
-  normalization, person inference, or psychometric estimation is permitted.
-- Reject divergent duplicate identities and owner-level truncation. Task
-  Ratings remain unavailable until their integer Task IDs and statements have
-  a separate normalized source-target contract.
-
-Acceptance: the replay-safe migration creates the normalized store; the pinned
-CSV importer validates both rating and scale-reference digests and row counts,
-reference identity, source scale, uncertainty, flags, and dates before
-persistence; PostgreSQL integration proves missing partitions fail closed and
-repeated null-category UPSERT is idempotent.
-API, UI, and derived modeling remain unavailable until separate accepted
-delivery records.
-
-### PRD-FR-2E — Occupation-rating evidence read
-
-- Let an authenticated user open one exact release/source/occupation profile
-  with both rating and scale artifact provenance (ADR 0258).
-- Distinguish an unavailable imported source from an available source with no
-  observation for the occupation.
-- Preserve exact decimal text, uncertainty, suppression, relevance, source
-  month, domain source, and declared bounds; derive no ranking or recommendation.
-
-Acceptance: invalid identifiers and unbounded pages are rejected; an unavailable
-source never appears as a negative profile; pagination is deterministic; and a
-suppressed observation retains its value and warning flag together.
-
-### PRD-FR-2F — Occupation-rating evidence view
-
-- Let an authenticated user submit an exact O*NET-SOC code, release, and source
-  from the existing Dashboard without changing the governed GNB (ADR 0259).
-- Display published values beside bounds, sample/error/interval evidence,
-  source time, and text warnings; link both source artifacts.
-- Give different next actions for unavailable source, empty occupation,
-  transport failure, and additional pages.
-
-Acceptance: keyboard users can operate the form and named horizontally
-scrollable table; narrow layouts retain complete values; suppression remains
-visible beside its value; and Storybook covers populated, narrow, unavailable,
-and empty states using synthetic data.
-
-### PRD-FR-2G — Imported rating-source catalog
-
-- Populate the occupation evidence selector only from imported artifacts that
-  contain observations, preserving release and artifact provenance (ADR 0260).
-- Exclude the scale-definition support artifact from the rating-source selector.
-- Disable profile submission and state the next action while the catalog is
-  loading, empty, or unavailable.
-
-Acceptance: a user never types an internal release/source code; the selector
-order follows persisted import time rather than parsed version heuristics; and
-the real PostgreSQL integration test proves an imported synthetic artifact is
-listed while its supporting scale artifact is not.
-
-### PRD-FR-2H — Occupations represented in a rating source
-
-- Populate the occupation selector with exact stored code/title pairs that
-  have observations in the selected imported source (ADR 0261).
-- Clear the current occupation and profile when the source changes, and clear
-  the profile when the occupation changes; never mix continuation rows across
-  occupations or sources.
-- Keep unavailable source, available-empty source, loading, and transport
-  failure distinct and actionable.
-
-Acceptance: a user selects a stored title rather than typing an internal code;
-the PostgreSQL integration test proves the source membership predicate; and
-component tests prove selector changes clear prior evidence and pagination
-stays bound to the loaded profile identifiers.
-
-### PRD-FR-2I — Occupation catalog title filter
-
-- Let an authenticated user filter the imported occupation catalog by
-  published title or retained code without ranking or typed-code fallback
-  (ADR 0262).
-- Reset the filter when the source changes.
-- Disable profile submission and state the next action when the filter
-  matches no catalog occupation.
-
-Acceptance: submitting still sends only a catalog identity; a non-matching
-filter never creates a request; and Storybook covers a no-match state.
-
-### PRD-FR-2J — Authorized job-family and job-series snapshots
-
-- Import one authorized, pinned organization-specific source snapshot without
-  committing runtime rows or creating an organization (ADR 0263).
-- Keep job families, job series, standard occupations, organizational units,
-  positions, people, and psychological constructs as distinct identities.
-- Preserve source-declared multiple-family membership and validity dates; infer
-  no parent or occupation binding from a code, label, similarity, or model.
-- Persist a standard-occupation binding only when scheme IRI, version, code,
-  and source relation are all explicitly supplied.
-
-Acceptance: synthetic tests reproduce a series with two source-declared family
-parents, reject cycles and partial bindings, leave an occupation-looking label
-unbound, and prove the normalized snapshot store is immutable.
-
 ### PRD-FR-3 — Bounded ontology exploration
 
 - Apply RBAC/ABAC, source eligibility, and knowledge cutoff before graph
diff --git a/docs/product-technical-gap-baseline.md b/docs/product-technical-gap-baseline.md
index dccd12f40..4f6a34faf 100644
--- a/docs/product-technical-gap-baseline.md
+++ b/docs/product-technical-gap-baseline.md
@@ -1,5 +1,67 @@
 # Product & Technical Gap Baseline
 
+> Exact-head overlay: 2026-08-28 00:21 KST. Protected `main` was
+> `96109cf06f6611eb0e09c43d33416f8f8cf4242f`; 10 PRs and 10 issues were
+> open. This overlay supersedes every older queue count and head snapshot in
+> this document. Historical sections remain provenance for earlier decisions,
+> not current release evidence.
+
+## Current protected-delivery and Voice-of-X decision
+
+ADRs 0246 and 0251 remain distinct authorities: ADR 0246 governs the twelve
+atomic, extensible Voice-of-X concepts and ADR 0251 governs the FJA
+I/O-psychology semantic layer. ADR 0256 governs evidence-bearing Voice
+composition. No fixed compound-code catalog, B2B2C-only cross-product,
+confidence weight, or keyword inference is authorized. The accepted literature
+and standards cited by those ADRs support an extensible stakeholder vocabulary;
+they do not prove that any finite set of combinations is exhaustive.
+
+Protected `main` implements the twelve atomic concepts, normalized additional
+Voice assignments, server-created PROV-O derivation, truth status, authorization
+filtering, separate carrying-Post and derivation-evidence actions, and paged
+JSON-LD union behavior. Those are implementation facts, not evidence that a
+particular runtime contains accepted assignments. Current non-identifying
+repository aggregates establish only the open-work inventory above; no
+production row, title, organization, or credential was inspected for this
+overlay.
+
+The largest remaining user-visible Voice gap is temporal imported-primary
+history. PR #752 exact head `5c2dfed5a32db655264fa62fa46fda626d753954`
+adds ADR 0252 and migration 0243 so recurring A to B to A changes retain
+non-overlapping half-open intervals. The review repair removes duplicate SQL
+clauses and makes ontology continuation evaluate the interval containing the
+frozen page snapshot rather than requiring the current open interval. Focused
+local Voice/ontology regression evidence passed 93 tests after merging current
+`main`; the full frontend suite passed 453 tests. The combined-Voice Storybook
+scene was rendered and visually inspected at 1440 by 1000 and 390 by 844 CSS
+pixels: carrying-Post and derivation-evidence controls remain distinct, and the
+mobile exact-value table preserves its horizontal-scroll access. Hosted
+exact-head checks were still pending. Authenticated PostgreSQL API and
+authenticated product-UI evidence have not been recollected for this head
+because the running canonical Compose project was assembled from multiple
+other active worktrees; it was not rebuilt or migrated over their work. Runtime
+acceptance therefore remains **unavailable**, not complete.
+
+| PR | Exact head | Current gate at overlay time |
+|---:|---|---|
+| #629 | `fcb933ba30a5fe866f01a5a7940c7304cce347f5` | Mergeable with auto-merge retained; exact-head checks/reviews pending and `CHANGES_REQUESTED` not cleared |
+| #640 | `bd73e0a43ae139d0091c25696e89786c62a34e74` | Conflicting with `main`; auto-merge retained |
+| #643 | `bae04cff19f2e1d54237b7d7168708b9c5df9a3b` | Draft and conflicting; no auto-merge |
+| #644 | `f53dd28e50f984ceebd9520f800f84dbea05b7e6` | Conflicting; review required; auto-merge retained |
+| #667 | `0c0f4af572a94e63cc8ea4545e48f5eda32a389c` | Conflicting with one prior failed check; review required; auto-merge retained |
+| #668 | `234f975ba5b0982e22d918219b505a9cd6a103e2` | Conflicting; review required; auto-merge retained |
+| #672 | `a3e87a89185fae03c5f18c79e2d97d12c73e8af9` | Conflicting; review required; auto-merge retained |
+| #679 | `135dfe7c4266c7a2098c622b7c9976eaf7304cdd` | Conflicting; review required; auto-merge retained |
+| #702 | `ebee9520a5b48c3074135153954c96e067309863` | Conflicting; no auto-merge |
+| #752 | `5c2dfed5a32db655264fa62fa46fda626d753954` | Mergeable; exact-head checks and independent review pending; no auto-merge yet |
+
+Organization ruleset 18156473 currently requires one approval, dismissal of
+stale approvals after a push, approval for unattributed changes, resolved
+threads, and seven workflows pinned to `.github@main`. Repository ruleset
+21065108 and the organization ruleset both prohibit non-fast-forward updates.
+No self-approval, admin bypass, force push, stale check transfer, or pre-parent
+stack evidence satisfies those gates.
+
 > Voice-of-X delivery snapshot: 2026-08-27 KST. Protected `main` was
 > `ff7431bd1851c03e737808d22c6a2d43968582f9`; PR #713 was
 > `850494c3861703862a76cfe564381a41243c6c2d`; stacked PR #717 was
@@ -620,6 +682,7 @@ this file per §3.5 of the prior snapshot).
 | Semantic source rendering | Paragraph, table, list, formula, and indentation work exists across stacks (#394, #427, #448–#450); #515 adds synthetic backend/frontend parity for deterministic rows/cells, footnote boundaries, and encoded scripts | Land the #427 → #515 stack, then gather authenticated browser evidence that list nesting, continuation alignment, and formula units render without authoring-layout artifacts |
 | Event and project semantics | #663 is the largest current user-visible gap slice: evidence-backed Project nodes, bounded traversal, cutoff/snapshot fencing, exact-value table parity, and localized graph labels. Focus visibility, label-bound, and temporal test-double regressions are repaired. #666's heuristic removal is composed into this parent but is not separately protected-main evidence. #640 separately adds project journeys without claiming authoritative lifecycle status | Combined #663 must pass exact-head checks and independent approval before protected merge. Aggregate authenticated evidence must still prove distinct projects/events and handover intervals without promoting co-occurrence |
 | Voice primary history | The #717 candidate updates ADR 0256 and migration 0237 with immutable assignment ids and half-open intervals, closing rather than deleting a replaced primary so A → B → A is representable; this is not protected-main evidence | Prove migration replay, concurrent primary changes, non-overlap, and API/ontology cutoff reads against synthetic PostgreSQL at the current exact head, then close #748 only after protected delivery |
+| Voice primary history | ADR 0252 and candidate migration 0243 retain half-open imported-primary intervals, allow recurring A → B → A values, and align live, cutoff, and ontology-snapshot reads. Static/unit contracts pass; this branch is not protected-main or authenticated runtime evidence | Prove migration replay and atomic A → B → A cutoff reads against the isolated PostgreSQL/OIDC stack, then pass exact-head hosted checks and protected review before closing #748 |
 | Knowledge Graph readability | #659 recreates the token-backed node-type repair on current `main`, including regression coverage; it is open and therefore not protected-main evidence | Merge #659 normally, then verify light/dark contrast, keyboard graph navigation, full labels, and evidence tables in the authenticated rendered surface |
 | Source-code lookup UX | Source state/detail codes remain evidence-bearing machine values and current detail presentation is dense | Catalog-backed display labels with raw-code provenance, compact 5W1H/source-detail hierarchy, keyboard access, and no unsupported customer/project binding |
 | Calendar / Naruon | #355 delivered the projection contract; v2.17.0 wires operator consumption without forwarding the end-user token. Naruon producer, provider/consumer fixtures, and protected merge remain open (#336) | Verify observed events against the published schema without invented events; keep commitments available when the channel is unwired |
@@ -781,23 +844,24 @@ post-merge reruns (not transferable evidence for later heads):
 | ---: | --- | --- |
 | #750 | Leftover-map unexplained leftover share persisted (`report_leftover_map_unexplained_share`, share `s = U² / R²`) | ADR 0233 |
 | #749 | Authorized job-family/job-series import snapshots (`0223_authorized_job_architecture`) | ADR 0263 |
-| #759 | ***Promoted** the ONET rating-store stack to `main`: migrations 0222/0223, authenticated rating/rating-sources/rating-occupations endpoints, `OccupationRatingProfile` UI + stories, rating client functions, import scripts, ADR 0252–0263 references. Semgrep SQLi nullified by PL/pgSQL `format(%I/%L)` DDL + documented `nosemgrep`; 1583 Python + 447 frontend tests green | ADR 0257–0263 |
 | #747 | Current product and MCP manuals (`docs/manuals/*`, contract tests) | ADR 0118-family |
 | #754 | Customer-actionable copy and ADR 0237 accelerator runtime boundary; share/bookmark/verification call sites reworded and ko/zh/ja/vi translations completed after review | ADR 0237 |
-| #700 | Source conversation-turn evidence ingestion (`0233_source_conversation_turn_evidence`, choke/adjacency resilience) | ADR 0238 |
-| #658 | Optional Global Ask knowledge cutoff honoring `source_post_revision` cover | ADR 0216 |
-| #632 | Graph-fact source provenance preserved through MCP streaming + verified psql-parity migration fixture | ADR 0166 |
 | #742 | Evidence-bound product-operations relations (stack base) | ADR 0235 |
 | #743 | Imported occupation-rating source catalog (stack base) | ADR 0260 |
 | #745 | Occupation catalog title filter (stack base) | ADR 0262 |
 | #746 | Rating-source occupation selector (stack base) | ADR 0261 |
 | #740 | Occupation rating evidence view (stack base) | ADR 0259 |
+| #732 | O*NET content-model published linkages (stack base) | ADR 0256 |
 | #720 | Cancel stale test runs on PR close | — |
 | #716 | Prioritized evidence-bound operations backfill | — |
 | #711 | Pinned validated structured-workflow runtime | — |
 | #704 | Current-main external lineage contract publication | — |
 
-The ONET rows stacked into base branches (#743/#745/#746/#740/#732) reached
-`main` together through the #759 promotion; their per-base merge records are
-historical evidence only. The job-architecture artifact ship originally via
-#749 is now re-verified on `main` from the promotion.
+Rebased and re-pushed onto current `main` (checks running at this snapshot):
+#700 source-conversation-turn contract (ADR 0238), #658 optional Global Ask
+knowledge cutoff (ADR 0216). Both unreviewed until exact-head checks pass.
+
+The ONET stack rows above landed into their stacked base branches rather than
+`main`; their content reaches `main` only if the base branch continues into a
+`main`-bound PR. Each base branch is recorded in the PR's `baseRefName` and
+remains the owner's responsibility to promote.
diff --git a/docs/voice-combination-technical-requirements.md b/docs/voice-combination-technical-requirements.md
new file mode 100644
index 000000000..f075f7e2e
--- /dev/null
+++ b/docs/voice-combination-technical-requirements.md
@@ -0,0 +1,60 @@
+# Voice-of-X Combination Technical Requirements
+
+This supporting TRD projects ADR 0246, ADR 0251, and ADR 0252. Those ADRs are
+normative when this document and an implementation differ.
+
+## Scope
+
+LineageWeave represents a Post's explicitly supplied stakeholder perspectives
+without assuming a company, B2B2C chain, or exhaustive industry taxonomy. One
+imported primary Voice and zero or more evidence-bearing additional Voices are
+atomic assignments; combinations are sets of rows, never compound codes.
+
+## Requirements
+
+| ID | Requirement | Verification |
+|---|---|---|
+| VOC-TR-1 | `source_post.voc_type_code` owns the imported primary; additional assignments cannot demote it | Database trigger and API conflict tests |
+| VOC-TR-2 | Every additional Voice references a normalized PROV-O derivation and governed truth status | Foreign keys, category trigger, authenticated write test |
+| VOC-TR-3 | Primary assignments use non-overlapping half-open effective intervals and allow A → B → A under serialized concurrent source updates | GiST exclusion constraint and PostgreSQL integration tests |
+| VOC-TR-4 | Live reads select current rows; cutoff reads select the containing interval; ontology continuation uses its frozen snapshot when no cutoff exists | Backend SQL-contract tests and authenticated cutoff API test |
+| VOC-TR-5 | Post, filter, ontology JSON-LD, exact-value CSV, and UI apply the same RBAC/ABAC and source-eligibility boundary | API, SHACL, frontend interaction, and accessibility tests |
+| VOC-TR-6 | Voice stays separate from counterparty relationship, role, topic, channel, lifecycle, and stakeholder salience | ADR/schema review and ontology round-trip tests |
+| VOC-TR-7 | Migration replay preserves existing starts and never reconstructs deleted pre-migration history | Migration replay test and non-identifying runtime evidence |
+
+## Read contract
+
+```text
+reference_time = knowledge_cutoff ?? ontology_snapshot ?? live
+live            = effective_to IS NULL
+historical      = effective_from <= reference_time < effective_to
+open historical = effective_from <= reference_time AND effective_to IS NULL
+```
+
+The interval is lower-inclusive and upper-exclusive. The exact primary-change
+instant belongs to the new primary, so a read cannot return two primary rows.
+
+## Component flow
+
+```mermaid
+flowchart LR
+  Import[Authorized source import] --> SourcePost[(source_post)]
+  SourcePost --> Trigger[Primary Voice sync trigger]
+  Trigger --> History[(source_post_voice intervals)]
+  Admin[post_admin + visible evidence] --> API[Voice assignment API]
+  API --> Provenance[(PROV-O assertion)]
+  Provenance --> History
+  History --> PostRead[Post and filters]
+  History --> Ontology[Ontology JSON-LD and CSV]
+  PostRead --> UI[Post and board UI]
+  Ontology --> Explorer[Ontology explorer]
+```
+
+## Failure behavior
+
+- Missing or hidden evidence rejects or omits the additional assignment; it is
+  never replaced with a placeholder.
+- Unknown Voice/truth categories fail with a database check error.
+- Overlapping imported-primary intervals fail at the database boundary.
+- Cutoffs before retained history return an explicit unavailable state rather
+  than the current value.
diff --git a/lineageweave/ontology_neighborhood.py b/lineageweave/ontology_neighborhood.py
index cb07daae0..991b81f10 100644
--- a/lineageweave/ontology_neighborhood.py
+++ b/lineageweave/ontology_neighborhood.py
@@ -234,7 +234,9 @@ class OntologyVoiceAssignment:
     is_primary: bool
     truth_status_code: str
     recorded_at: datetime
+    effective_from: datetime
     provenance_reference: str
+    effective_to: datetime | None = None
     evidence_post_id: str | None = None
 
     def __post_init__(self) -> None:
@@ -260,6 +262,16 @@ def __post_init__(self) -> None:
             raise OntologyNeighborhoodError(
                 "naive_timestamp", "voice assignment recorded_at must be offset-aware"
             )
+        if self.effective_from.tzinfo is None or (
+            self.effective_to is not None and self.effective_to.tzinfo is None
+        ):
+            raise OntologyNeighborhoodError(
+                "naive_timestamp", "voice assignment effective bounds must be offset-aware"
+            )
+        if self.effective_to is not None and self.effective_from >= self.effective_to:
+            raise OntologyNeighborhoodError(
+                "invalid_interval", "voice assignment effective interval is empty or inverted"
+            )
 
 
 @dataclass(frozen=True)
@@ -324,8 +336,10 @@ def exact_value_rows(self) -> tuple[dict[str, str], ...]:
                     "target_type_code": "node_voice_type",
                     "truth_status_code": assignment.truth_status_code,
                     "recorded_at": assignment.recorded_at.isoformat(),
-                    "valid_from": "",
-                    "valid_to": "",
+                    "valid_from": assignment.effective_from.isoformat(),
+                    "valid_to": assignment.effective_to.isoformat()
+                    if assignment.effective_to
+                    else "",
                     "evidence_count": "1" if assignment.evidence_post_id or assignment.is_primary else "0",
                     "evidence_post_id": assignment.evidence_post_id or (
                         assignment.post_id if assignment.is_primary else ""
@@ -414,23 +428,24 @@ def jsonld_document(self) -> dict[str, object]:
                 if evidence_iri is not None
                 else {}
             )
-            graph.append(
-                {
-                    "@id": assignment_iri,
-                    "@type": str(LW.VoiceAssignment),
-                    str(LW.assignedVoiceType): {"@id": assignment.voice_type_iri},
-                    str(LW.primaryVoiceAssignment): {
-                        "@value": assignment.is_primary,
-                        "@type": "xsd:boolean",
-                    },
-                    **provenance,
-                    "lw:truthStatus": assignment.truth_status_code,
-                    "prov:generatedAtTime": {
-                        "@value": assignment.recorded_at.isoformat(),
-                        "@type": "xsd:dateTimeStamp",
-                    },
-                }
+            item: dict[str, object] = {
+                "@id": assignment_iri,
+                "@type": str(LW.VoiceAssignment),
+                str(LW.assignedVoiceType): {"@id": assignment.voice_type_iri},
+                str(LW.primaryVoiceAssignment): {
+                    "@value": assignment.is_primary,
+                    "@type": "xsd:boolean",
+                },
+                **provenance,
+                "lw:truthStatus": assignment.truth_status_code,
+            }
+            _add_jsonld_times(
+                item,
+                assignment.recorded_at,
+                assignment.effective_from,
+                assignment.effective_to,
             )
+            graph.append(item)
             graph.append(
                 {
                     "@id": assignment.voice_type_iri,
diff --git a/migrations/0237_source_post_voice_combination.sql b/migrations/0237_source_post_voice_combination.sql
index a010a830c..5011dc1b3 100644
--- a/migrations/0237_source_post_voice_combination.sql
+++ b/migrations/0237_source_post_voice_combination.sql
@@ -129,7 +129,8 @@ update source_post_voice voice
 
 insert into source_post_voice
     (post_id, voice_type_code, is_primary, truth_status_code, effective_from)
-select post.post_id, post.voc_type_code, true, 'truth_observed', post.created_at
+select post.post_id, post.voc_type_code, true, 'truth_observed',
+       least(post.created_at, clock_timestamp())
   from source_post post
  where not exists (
     select 1 from source_post_voice voice
@@ -159,7 +160,7 @@ begin
         new.voc_type_code,
         true,
         'truth_observed',
-        case when tg_op = 'INSERT' then new.created_at else change_at end
+        case when tg_op = 'INSERT' then least(new.created_at, change_at) else change_at end
     )
     on conflict (post_id, voice_type_code) where effective_to is null do update
     set is_primary = true,
diff --git a/migrations/0243_source_post_voice_history.sql b/migrations/0243_source_post_voice_history.sql
new file mode 100644
index 000000000..ad4637754
--- /dev/null
+++ b/migrations/0243_source_post_voice_history.sql
@@ -0,0 +1,57 @@
+-- ADR 0252: preserve non-overlapping imported primary Voice intervals.
+
+begin;
+
+create extension if not exists btree_gist;
+
+alter table source_post_voice
+    add column if not exists effective_to timestamptz;
+
+alter table source_post_voice
+    drop constraint if exists source_post_voice_effective_interval_check;
+alter table source_post_voice
+    add constraint source_post_voice_effective_interval_check
+    check (effective_to is null or effective_from < effective_to);
+
+create unique index if not exists source_post_voice_current_pair_idx
+    on source_post_voice (post_id, voice_type_code)
+    where effective_to is null;
+
+alter table source_post_voice
+    drop constraint if exists source_post_voice_primary_period_excl;
+alter table source_post_voice
+    add constraint source_post_voice_primary_period_excl
+    exclude using gist (
+        post_id with =,
+        tstzrange(effective_from, effective_to, '[)') with &&
+    ) where (is_primary);
+
+create or replace function synchronize_source_post_primary_voice()
+returns trigger
+language plpgsql
+as $$
+declare
+    change_at timestamptz := clock_timestamp();
+begin
+    update source_post_voice
+       set effective_to = change_at
+     where post_id = new.post_id
+       and effective_to is null
+       and (is_primary or voice_type_code = new.voc_type_code);
+
+    insert into source_post_voice
+        (post_id, voice_type_code, is_primary, truth_status_code,
+         effective_from, recorded_at)
+    values (
+        new.post_id,
+        new.voc_type_code,
+        true,
+        'truth_observed',
+        case when tg_op = 'INSERT' then least(new.created_at, change_at) else change_at end,
+        change_at
+    );
+    return new;
+end;
+$$;
+
+commit;
diff --git a/tests/test_ontology_neighborhood.py b/tests/test_ontology_neighborhood.py
index 3f6165745..764bceaec 100644
--- a/tests/test_ontology_neighborhood.py
+++ b/tests/test_ontology_neighborhood.py
@@ -912,6 +912,7 @@ def test_voice_assignments_join_exact_csv_rows_and_jsonld() -> None:
         is_primary=False,
         truth_status_code=TRUTH_OBSERVED,
         recorded_at=T0,
+        effective_from=T0,
         provenance_reference="Evidence-backed additional voice",
         evidence_post_id=POST_ID,
     )
@@ -922,6 +923,7 @@ def test_voice_assignments_join_exact_csv_rows_and_jsonld() -> None:
     assert row["target_label"] == "Voice of Process"
     assert row["evidence_post_id"] == POST_ID
     assert row["evidence_count"] == "1"
+    assert row["valid_from"] == T0.isoformat()
     graph = neighborhood.jsonld_document()["@graph"]
     assignment_iri = str(LW[f"voice-assignment/{POST_ID}/vops"])
     projected = next(item for item in graph if item.get("@id") == assignment_iri)
@@ -939,6 +941,10 @@ def test_voice_assignments_join_exact_csv_rows_and_jsonld() -> None:
     assert projected["prov:wasDerivedFrom"] == {
         "@id": ontology_node_iri(NODE_POST, POST_ID)
     }
+    assert (
+        projected["time:hasBeginning"]["time:inXSDDateTimeStamp"]["@value"]
+        == T0.isoformat()
+    )
 
     hidden_evidence = replace(assignment, evidence_post_id=None)
     hidden_row = replace(
@@ -957,6 +963,8 @@ def test_voice_assignments_join_exact_csv_rows_and_jsonld() -> None:
 
     with pytest.raises(OntologyNeighborhoodError, match="offset-aware"):
         replace(assignment, recorded_at=T0.replace(tzinfo=None))
+    with pytest.raises(OntologyNeighborhoodError, match="offset-aware"):
+        replace(assignment, effective_from=T0.replace(tzinfo=None))
 
 
 def test_node_bound_truncation_keeps_nearer_hop_over_farther_alphabetically_earlier_type() -> None:
diff --git a/tests/test_ontology_neighborhood_ingestion.py b/tests/test_ontology_neighborhood_ingestion.py
index 333d91df9..546fd28bb 100644
--- a/tests/test_ontology_neighborhood_ingestion.py
+++ b/tests/test_ontology_neighborhood_ingestion.py
@@ -930,6 +930,8 @@ def test_focus_label_fetch_may_be_empty_when_facts_already_labeled() -> None:
                 "is_primary": True,
                 "truth_status_code": "truth_observed",
                 "recorded_at": T0,
+                "effective_from": T0,
+                "effective_to": None,
                 "has_assertion": False,
                 "evidence_post_id": None,
             }
@@ -1003,6 +1005,8 @@ def test_load_voice_assignments_preserves_truth_and_customer_safe_provenance() -
                     "is_primary": True,
                     "truth_status_code": "truth_observed",
                     "recorded_at": T0,
+                    "effective_from": T0,
+                    "effective_to": None,
                     "has_assertion": False,
                     "evidence_post_id": None,
                 },
@@ -1013,6 +1017,8 @@ def test_load_voice_assignments_preserves_truth_and_customer_safe_provenance() -
                     "is_primary": False,
                     "truth_status_code": "truth_observed",
                     "recorded_at": T0,
+                    "effective_from": T0,
+                    "effective_to": None,
                     "has_assertion": True,
                     "evidence_post_id": POST_ID,
                 },
@@ -1030,7 +1036,15 @@ def test_load_voice_assignments_preserves_truth_and_customer_safe_provenance() -
     assert assignments[1].evidence_post_id == POST_ID
     assert "evidence.node_id = any($1::uuid[])" in conn.calls[0][0]
     assert "voice.is_primary or evidence.node_id = any($1::uuid[])" in conn.calls[0][0]
-    assert "voice.effective_from <= $2" in conn.calls[0][0]
+    assert (
+        "voice.effective_from <= coalesce($2::timestamptz, $3::timestamptz)"
+        in conn.calls[0][0]
+    )
+    assert (
+        "coalesce($2::timestamptz, $3::timestamptz) < voice.effective_to"
+        in conn.calls[0][0]
+    )
+    assert "$2::timestamptz is null and voice.effective_to is null" not in conn.calls[0][0]
     assert "voice.recorded_at <= $3" in conn.calls[0][0]
     assert conn.calls[0][1] == ([POST_ID], T0, T0)
 
diff --git a/tests/test_post_filter_options.py b/tests/test_post_filter_options.py
index 6ac4dc570..607a9f1bd 100644
--- a/tests/test_post_filter_options.py
+++ b/tests/test_post_filter_options.py
@@ -57,6 +57,7 @@ def test_post_filter_options_use_one_authorized_source_scan() -> None:
     assert "cross join lateral" in query
     assert "('post_visibility', post.visibility_code)" in query
     assert "left join source_post_voice voice" in query
+    assert query.count("voice.effective_to is null") == 1
     assert "('voc_type', coalesce(voice.voice_type_code, post.voc_type_code))" in query
     assert "post.corporate_entity_id::text = any($1::text[])" in query
     assert "post.process_unit_id::text = any($2::text[])" in query
diff --git a/tests/test_source_post_voice_history_schema.py b/tests/test_source_post_voice_history_schema.py
new file mode 100644
index 000000000..66b7866b9
--- /dev/null
+++ b/tests/test_source_post_voice_history_schema.py
@@ -0,0 +1,49 @@
+"""Static contract tests for ADR 0252 temporal primary Voice history."""
+
+from __future__ import annotations
+
+from pathlib import Path
+
+MIGRATION = (
+    Path(__file__).resolve().parents[1]
+    / "migrations"
+    / "0243_source_post_voice_history.sql"
+)
+
+
+def test_primary_voice_history_uses_non_overlapping_half_open_intervals() -> None:
+    """The database preserves recurring Voices and rejects overlapping primaries."""
+    sql = MIGRATION.read_text(encoding="utf-8").lower()
+
+    assert "add column if not exists effective_to timestamptz" in sql
+    assert "drop constraint if exists source_post_voice_pkey" not in sql
+    assert "primary key (post_id, voice_type_code, effective_from)" not in sql
+    assert "voice_assignment_id" not in sql
+    assert "tstzrange(effective_from, effective_to, '[)') with &&" in sql
+    assert "where (is_primary)" in sql
+    assert "effective_from < effective_to" in sql
+
+
+def test_primary_voice_change_closes_current_rows_before_insert() -> None:
+    """One transaction instant closes the prior state and opens the new primary."""
+    sql = MIGRATION.read_text(encoding="utf-8").lower()
+
+    assert "change_at timestamptz := clock_timestamp()" in sql
+    assert "set effective_to = change_at" in sql
+    assert "and (is_primary or voice_type_code = new.voc_type_code)" in sql
+    assert sql.index("set effective_to = change_at") < sql.index(
+        "insert into source_post_voice"
+    )
+    assert "on conflict" not in sql
+
+
+def test_future_source_clock_is_bounded_by_the_recording_clock() -> None:
+    """A future source timestamp cannot create an interval that closes backwards."""
+    assignment_sql = (
+        MIGRATION.parent / "0237_source_post_voice_combination.sql"
+    ).read_text(encoding="utf-8").lower()
+    history_sql = MIGRATION.read_text(encoding="utf-8").lower()
+
+    assert "least(post.created_at, clock_timestamp())" in assignment_sql
+    assert "least(new.created_at, change_at)" in assignment_sql
+    assert "least(new.created_at, change_at)" in history_sql
diff --git a/tests/test_source_post_voice_ingestion.py b/tests/test_source_post_voice_ingestion.py
index 75a0c48d0..52fdf0b1f 100644
--- a/tests/test_source_post_voice_ingestion.py
+++ b/tests/test_source_post_voice_ingestion.py
@@ -72,6 +72,7 @@ def test_additional_voice_creates_prov_derivation_and_assignment_atomically() ->
     assert "prov_was_derived_from" in sql
     assert "where effective_to is null" in sql
     assert "where not source_post_voice.is_primary" in sql
+    assert "where effective_to is null" in sql
     assert "voice-assignment/aaaaaaaa-aaaa-aaaa-aaaa-aaaaaaaaaaa1/vops" in str(
         conn.calls
     )
diff --git a/tests/test_source_post_voice_schema.py b/tests/test_source_post_voice_schema.py
index 05265bf10..b78135b6b 100644
--- a/tests/test_source_post_voice_schema.py
+++ b/tests/test_source_post_voice_schema.py
@@ -25,9 +25,10 @@ def test_voice_combination_schema_is_normalized_and_evidence_bearing() -> None:
     assert "effective_from timestamptz not null" in sql
     assert "true, 'truth_observed'" in sql
     assert "where is_primary" in sql
-    assert "select post.post_id, post.voc_type_code, true, 'truth_observed', post.created_at" in sql
+    assert "select post.post_id, post.voc_type_code, true, 'truth_observed'" in sql
+    assert "least(post.created_at, clock_timestamp())" in sql
     assert "change_at timestamptz := clock_timestamp()" in sql
-    assert "case when tg_op = 'insert' then new.created_at else change_at end" in sql
+    assert "case when tg_op = 'insert' then least(new.created_at, change_at) else change_at end" in sql
     assert "after insert on source_post" in sql
     assert "after update of voc_type_code on source_post" in sql
     assert "when (old.voc_type_code is distinct from new.voc_type_code)" in sql
diff --git a/tests/test_source_state_serialization.py b/tests/test_source_state_serialization.py
index 661c4e2c5..1153ac6e4 100644
--- a/tests/test_source_state_serialization.py
+++ b/tests/test_source_state_serialization.py
@@ -80,6 +80,7 @@ async def fetch(
             assert "provenance_assertion_id is not null as evidence_available" in query
             assert "voice.effective_from <= $2" in query
             assert "$2 < voice.effective_to" in query
+            assert "voice.effective_to is null or $2 < voice.effective_to" in query
             assert post_id == "aaaaaaaa-aaaa-aaaa-aaaa-aaaaaaaaaaaa"
             assert effective_cutoff == datetime(2026, 1, 1, tzinfo=UTC)
             return [