feat(lineage): Phase 1 backend — dependents endpoints + REGRESSION test

dict-relationships-v2 epic, Phase 1 read-side. Backend готов; admin UI
(card "Used by" + record references) — отдельным commit'ом.

Что добавлено:
- `DependentsQuery` (interface) + `DependentsQueryJdbc` (impl) — native
  SQL для record-level dependents lookup (paginated). Interface extracted
  чтобы Mockito мог mock'нуть на JDK 25 (concrete-class inline-mock incompat).
- `LineageIndexService` — два режима:
  * `findSchemaDependents(name, scopes)` — schema-level reverse FK list
    (in-memory traversal definitions, microseconds). Skip self-reference,
    scope-hidden sources, malformed x-references (logged warn).
  * `findRecordDependents(dict, businessKey, scopes, page, size)` —
    record-level paginated, perSource summary + total. Phase 2 переносит
    на materialized view; пока живём с прямым JSONB query (acceptable
    для <5k SLO).
- `LineageController` — два endpoint'а:
  * `GET /api/v1/dictionaries/{name}/dependents`
  * `GET /api/v1/records/{dict}/{businessKey}/dependents?page&size`
  Errors: 404 dictionary_not_found / record_not_active, 400 invalid_pagination.
  Scope hiding делает target invisible (empty list / 404 — без leak existence).

Tests:
- LineageIndexServiceTest (16 unit tests):
  * Schema-level: not_found, scope hidden, single ref + count, onClose
    propagation (BLOCK default + CASCADE explicit), sorting, source scope
    hidden, self-reference skip, malformed skip, target without properties.
  * Record-level: invalid pagination (page<0, size<1, size>200),
    dict_not_found, record_not_active, no schema refs → empty page,
    1 source с dependents → items+perSource+total, pagination через
    page=0/1/2 size=2 (3 страницы из 5 records).
- ReferenceCloseRegressionTest (4 mandatory regression tests):
  * v1.2.1 schema без x-references-on-close — validate() возвращает
    те же error codes (никаких новых on_close/cascade в codes).
  * Schema parser default'ит onClose=BLOCK для v1.2.1 schemas.
  * Legacy single-arg parseRef(spec) без property schema сохраняет contract.
  * Schema без x-references вообще — full no-op (30+ legacy справочников).

mvn -pl ordinis-rest-api -am test: 89/89 PASS (5.686s).
mvn verify -pl '!ordinis-app': все модули SUCCESS (9.031s).

Backward-compat: bundle v1.2.1 → identical поведение к v1.0.7. Cascade engine
не поднимается на closed-side (Phase 3 territory).

ordinis-domain/src/main/java/cloud/nstart/terravault/ordinis/domain/record/DependentsQuery.java

@@ -0,0 +1,72 @@
+package cloud.nstart.terravault.ordinis.domain.record;
+
+import java.time.OffsetDateTime;
+import java.util.List;
+import java.util.UUID;
+
+/**
+ * Native SQL для record-level dependents lookup: "какие active записи в
+ * source dictionary ссылаются на target record через JSONB
+ * {@code data->>source_field == target_value}".
+ *
+ * <p>Используется {@code LineageIndexService.findRecordDependents} в Phase 1.
+ * Phase 2 переносит этот lookup на materialized view с UNIQUE INDEX(id) для
+ * REFRESH CONCURRENTLY — пока живём с прямым JSONB query (acceptable для
+ * {@code <}5k records SLO).
+ *
+ * <p>Interface вместо concrete класса — JDK 25 + Mockito inline-mock не работает
+ * на concrete classes (см. memo "JDK 25 + Mockito subclass mock-maker incompat").
+ * Default impl {@link DependentsQueryJdbc} автоинджектится через @Component.
+ */
+public interface DependentsQuery {
+
+  /**
+   * Active records в source_dict где {@code data->>source_field == value}.
+   * Возвращает paginated slice + total count (один call для UI: count first,
+   * потом page если total > 0).
+   *
+   * <p>NULL-значения в source_field автоматически исключены SQL'ом
+   * (NULL != value никогда не true).
+   *
+   * <p>Order by created_at DESC чтобы newest dependents видны первыми
+   * (admin debug pattern).
+   *
+   * @param sourceDictionaryId source dict (referencing)
+   * @param sourceField        JSONB key в source data что хранит ref value
+   * @param targetValue        фактическое значение target field в target record
+   * @param at                 момент времени — обычно now()
+   * @param limit              page size (caller должен ограничить ≤ 200)
+   * @param offset             page offset (≥ 0)
+   * @return slice records sorted by created_at DESC
+   */
+  List<DependentRow> findActive(
+      UUID sourceDictionaryId,
+      String sourceField,
+      String targetValue,
+      OffsetDateTime at,
+      int limit,
+      int offset);
+
+  /**
+   * Total count active dependents — отдельный query без LIMIT/OFFSET для
+   * pagination header. Faster чем {@code SELECT COUNT(*)} над основным
+   * results когда LIMIT/OFFSET идёт в отдельном request.
+   */
+  long countActive(
+      UUID sourceDictionaryId,
+      String sourceField,
+      String targetValue,
+      OffsetDateTime at);
+
+  /**
+   * Lightweight projection — лишний JSONB data в payload не нужен на этом
+   * endpoint, экономим bandwidth.
+   */
+  record DependentRow(
+      UUID id,
+      UUID dictionaryId,
+      String businessKey,
+      OffsetDateTime validFrom,
+      OffsetDateTime validTo,
+      OffsetDateTime createdAt) {}
+}

ordinis-domain/src/main/java/cloud/nstart/terravault/ordinis/domain/record/DependentsQueryJdbc.java

@@ -0,0 +1,97 @@
+package cloud.nstart.terravault.ordinis.domain.record;
+
+import org.springframework.beans.factory.annotation.Autowired;
+import org.springframework.jdbc.core.JdbcTemplate;
+import org.springframework.stereotype.Component;
+
+import java.time.OffsetDateTime;
+import java.util.List;
+import java.util.UUID;
+
+/**
+ * JdbcTemplate-backed default impl {@link DependentsQuery}.
+ *
+ * <p>Поля {@code source_field} приходят из schema parser (валидируются на
+ * bundle import), плюс whitelist regex здесь — SQL injection невозможна.
+ * Native SQL потому что JdbcTemplate не bind'ит JSONB key paths.
+ */
+@Component
+public class DependentsQueryJdbc implements DependentsQuery {
+
+  private final JdbcTemplate jdbc;
+
+  @Autowired
+  public DependentsQueryJdbc(JdbcTemplate jdbc) {
+    this.jdbc = jdbc;
+  }
+
+  @Override
+  public List<DependentRow> findActive(
+      UUID sourceDictionaryId,
+      String sourceField,
+      String targetValue,
+      OffsetDateTime at,
+      int limit,
+      int offset) {
+    validateField(sourceField);
+    if (limit <= 0 || limit > 500) {
+      throw new IllegalArgumentException("limit must be 1..500, got " + limit);
+    }
+    if (offset < 0) {
+      throw new IllegalArgumentException("offset must be >= 0, got " + offset);
+    }
+
+    String sql = """
+        SELECT id, dictionary_id, business_key, valid_from, valid_to, created_at
+        FROM dictionary_records
+        WHERE dictionary_id = ?
+          AND valid_from <= ?
+          AND valid_to > ?
+          AND data ->> '%s' = ?
+        ORDER BY created_at DESC
+        LIMIT ? OFFSET ?
+        """.formatted(sourceField);
+
+    return jdbc.query(
+        sql,
+        (rs, n) -> new DependentRow(
+            (UUID) rs.getObject("id"),
+            (UUID) rs.getObject("dictionary_id"),
+            rs.getString("business_key"),
+            rs.getObject("valid_from", OffsetDateTime.class),
+            rs.getObject("valid_to", OffsetDateTime.class),
+            rs.getObject("created_at", OffsetDateTime.class)),
+        sourceDictionaryId, at, at, targetValue, limit, offset);
+  }
+
+  @Override
+  public long countActive(
+      UUID sourceDictionaryId,
+      String sourceField,
+      String targetValue,
+      OffsetDateTime at) {
+    validateField(sourceField);
+
+    String sql = """
+        SELECT COUNT(*) FROM dictionary_records
+        WHERE dictionary_id = ?
+          AND valid_from <= ?
+          AND valid_to > ?
+          AND data ->> '%s' = ?
+        """.formatted(sourceField);
+
+    Long count = jdbc.queryForObject(sql, Long.class, sourceDictionaryId, at, at, targetValue);
+    return count == null ? 0 : count;
+  }
+
+  /** Paranoid SQL injection guard. Mirror'ит {@code OrphanReferenceQuery}. */
+  private static void validateField(String field) {
+    if (field == null || field.isBlank()) {
+      throw new IllegalArgumentException("field name is empty");
+    }
+    if (!field.matches("^[a-zA-Z_][a-zA-Z0-9_]{0,127}$")) {
+      throw new IllegalArgumentException(
+          "field name has unsafe characters: " + field);
+    }
+  }
+}

ordinis-rest-api/src/main/java/cloud/nstart/terravault/ordinis/restapi/service/reference/LineageIndexService.java

@@ -0,0 +1,289 @@
+package cloud.nstart.terravault.ordinis.restapi.service.reference;
+
+import cloud.nstart.terravault.ordinis.domain.DataScope;
+import cloud.nstart.terravault.ordinis.domain.definition.DictionaryDefinition;
+import cloud.nstart.terravault.ordinis.domain.definition.DictionaryDefinitionRepository;
+import cloud.nstart.terravault.ordinis.domain.record.DependentsQuery;
+import cloud.nstart.terravault.ordinis.domain.record.DictionaryRecord;
+import cloud.nstart.terravault.ordinis.domain.record.DictionaryRecordRepository;
+import com.fasterxml.jackson.databind.JsonNode;
+import org.slf4j.Logger;
+import org.slf4j.LoggerFactory;
+import org.springframework.stereotype.Service;
+import org.springframework.transaction.annotation.Transactional;
+
+import java.time.OffsetDateTime;
+import java.util.ArrayList;
+import java.util.HashMap;
+import java.util.Iterator;
+import java.util.List;
+import java.util.Map;
+import java.util.Optional;
+import java.util.Set;
+
+/**
+ * Reverse FK lineage index: "кто ссылается на этот словарь / запись".
+ *
+ * <p>Phase 1 read-side для dict-relationships-v2 epic. Два режима:
+ *
+ * <ul>
+ *   <li><b>Schema-level</b>: {@link #findSchemaDependents(String, Set)} —
+ *       берёт target dict name, обходит все definitions и собирает все
+ *       fields у которых {@code x-references = "<target>.<field>"}. Дёшево
+ *       (in-memory traversal, ~40 dicts × ~20 fields), кэш не нужен в v1.</li>
+ *   <li><b>Record-level</b>: {@link #findRecordDependents} — для конкретного
+ *       business_key в target_dict находит active records в source dicts
+ *       что на него ссылаются. Paginated через {@link DependentsQuery} (native
+ *       SQL JSONB lookup). Phase 2 переносит на materialized view.</li>
+ * </ul>
+ *
+ * <p>Caching: Phase 1 не кэширует — schema lookup микросекунды, record-level
+ * SQL paginated и каждый bbox разный. Phase 2 добавит {@code @Cacheable} для
+ * schema-level + invalidation на bundle import (см. eng review A.1).
+ *
+ * <p>Scope: filter target dict + source dicts через {@code allowedScopes}.
+ * Если каллер не имеет access к target dict — возвращаем "not found"
+ * (consistent с CRUD endpoint behaviour, чтобы не leak'ало existence).
+ */
+@Service
+public class LineageIndexService {
+
+  private static final Logger log = LoggerFactory.getLogger(LineageIndexService.class);
+
+  private final DictionaryDefinitionRepository definitionRepository;
+  private final DictionaryRecordRepository recordRepository;
+  private final DependentsQuery dependentsQuery;
+
+  public LineageIndexService(
+      DictionaryDefinitionRepository definitionRepository,
+      DictionaryRecordRepository recordRepository,
+      DependentsQuery dependentsQuery) {
+    this.definitionRepository = definitionRepository;
+    this.recordRepository = recordRepository;
+    this.dependentsQuery = dependentsQuery;
+  }
+
+  /**
+   * Schema-level dependents target dict: какие dict.field ссылаются.
+   *
+   * @param targetDictName  dict name (e.g. "satellite_type")
+   * @param allowedScopes   что caller видит (источники с недоступным scope filtered)
+   * @return list reverse FK pointers, sorted by sourceDict, sourceField asc.
+   *         Empty если target не существует или у caller нет к нему access.
+   *
+   * @throws SchemaDependentsTarget#NOT_FOUND нет такого dict.
+   */
+  @Transactional(readOnly = true)
+  public List<SchemaDependent> findSchemaDependents(
+      String targetDictName, Set<DataScope> allowedScopes) {
+    Optional<DictionaryDefinition> targetOpt = definitionRepository.findByName(targetDictName);
+    if (targetOpt.isEmpty()) return List.of();
+    DictionaryDefinition target = targetOpt.get();
+    if (!allowedScopes.contains(target.getScope())) {
+      // Scope hide — target invisible. Return empty (same shape as not_found).
+      return List.of();
+    }
+
+    List<DictionaryDefinition> all = definitionRepository.findAll();
+    // Pre-compute counts active records per source dict (для UI badge).
+    Map<java.util.UUID, Long> counts = new HashMap<>();
+    OffsetDateTime now = OffsetDateTime.now();
+    for (Object[] row : recordRepository.countActiveGroupedByDictionary(allowedScopes, now)) {
+      counts.put((java.util.UUID) row[0], (Long) row[1]);
+    }
+
+    List<SchemaDependent> out = new ArrayList<>();
+    for (DictionaryDefinition source : all) {
+      if (source.getId().equals(target.getId())) continue; // self-reference skipped
+      if (!allowedScopes.contains(source.getScope())) continue;
+      JsonNode schema = source.getSchemaJson();
+      if (schema == null || !schema.isObject()) continue;
+      JsonNode properties = schema.get("properties");
+      if (properties == null || !properties.isObject()) continue;
+
+      Iterator<Map.Entry<String, JsonNode>> it = properties.fields();
+      while (it.hasNext()) {
+        Map.Entry<String, JsonNode> entry = it.next();
+        String fieldName = entry.getKey();
+        JsonNode propSchema = entry.getValue();
+        JsonNode refNode = propSchema.get("x-references");
+        if (refNode == null || !refNode.isTextual()) continue;
+
+        ReferenceValidator.ParsedRef parsed;
+        try {
+          parsed = ReferenceValidator.parseRef(refNode.asText(), propSchema);
+        } catch (IllegalArgumentException e) {
+          log.warn("Schema {} field {} malformed x-references: {}",
+              source.getName(), fieldName, e.getMessage());
+          continue;
+        }
+        if (!parsed.dictionaryName().equals(target.getName())) continue;
+
+        out.add(new SchemaDependent(
+            source.getName(),
+            source.getDisplayName(),
+            fieldName,
+            parsed.fieldName(),
+            parsed.onClose(),
+            counts.getOrDefault(source.getId(), 0L)));
+      }
+    }
+    out.sort((a, b) -> {
+      int c = a.sourceDict().compareTo(b.sourceDict());
+      return c != 0 ? c : a.sourceField().compareTo(b.sourceField());
+    });
+    return out;
+  }
+
+  /**
+   * Record-level dependents: для target record (dict + business_key) ищет
+   * все active records в source dicts что на него ссылаются.
+   *
+   * <p>Логика: для каждого source dict у которого schema имеет
+   * {@code x-references: "<target_dict>.<target_field>"} — берём value
+   * {@code target_record.data->>target_field} и считаем/перечисляем active
+   * source records где {@code data->>source_field == that_value}.
+   *
+   * @param page page index 0..N
+   * @param size page size 1..200
+   * @return paged dependents + per-source totals
+   */
+  @Transactional(readOnly = true)
+  public RecordDependentsPage findRecordDependents(
+      String targetDictName,
+      String targetBusinessKey,
+      Set<DataScope> allowedScopes,
+      int page,
+      int size) {
+    if (page < 0) throw new IllegalArgumentException("page must be >= 0");
+    if (size < 1 || size > 200) {
+      throw new IllegalArgumentException("size must be 1..200, got " + size);
+    }
+
+    Optional<DictionaryDefinition> targetOpt = definitionRepository.findByName(targetDictName);
+    if (targetOpt.isEmpty()) {
+      throw new IllegalArgumentException("dictionary not found: " + targetDictName);
+    }
+    DictionaryDefinition target = targetOpt.get();
+    if (!allowedScopes.contains(target.getScope())) {
+      throw new IllegalArgumentException("dictionary not found: " + targetDictName);
+    }
+
+    OffsetDateTime now = OffsetDateTime.now();
+    Optional<DictionaryRecord> recOpt =
+        recordRepository.findActiveAt(target.getId(), targetBusinessKey, now);
+    if (recOpt.isEmpty()) {
+      throw new IllegalArgumentException(
+          "active record not found: " + targetDictName + "/" + targetBusinessKey);
+    }
+    DictionaryRecord targetRecord = recOpt.get();
+
+    // Schema-level reverse index: какие (source_dict, source_field, target_field, onClose).
+    List<SchemaDependent> schemaRefs = findSchemaDependents(targetDictName, allowedScopes);
+    if (schemaRefs.isEmpty()) {
+      return new RecordDependentsPage(
+          targetDictName, targetBusinessKey, List.of(), List.of(), 0, page, size);
+    }
+
+    // Per-source: pull target value, count + paginate.
+    // Phase 1: linear scan по source schemas — ОК для <40 dicts.
+    // Phase 2: MV сворачивает в один query.
+    List<RecordDependent> all = new ArrayList<>();
+    List<PerSourceSummary> summary = new ArrayList<>();
+    long total = 0;
+
+    for (SchemaDependent ref : schemaRefs) {
+      JsonNode targetData = targetRecord.getData();
+      if (targetData == null) continue;
+      JsonNode valueNode = targetData.get(ref.targetField());
+      if (valueNode == null || !valueNode.isTextual()) continue;
+      String targetValue = valueNode.asText();
+
+      Optional<DictionaryDefinition> sourceOpt =
+          definitionRepository.findByName(ref.sourceDict());
+      if (sourceOpt.isEmpty()) continue;
+      DictionaryDefinition sourceDef = sourceOpt.get();
+
+      long perSourceCount = dependentsQuery.countActive(
+          sourceDef.getId(), ref.sourceField(), targetValue, now);
+      summary.add(new PerSourceSummary(
+          ref.sourceDict(),
+          ref.sourceDisplayName(),
+          ref.sourceField(),
+          ref.onClose(),
+          perSourceCount));
+      total += perSourceCount;
+
+      if (perSourceCount == 0) continue;
+
+      // Paginate by aggregated total: каждый source pulled полностью если
+      // нужен (proper offset с union в SQL отложен Phase 2 — здесь simpler).
+      List<DependentsQuery.DependentRow> rows = dependentsQuery.findActive(
+          sourceDef.getId(), ref.sourceField(), targetValue, now,
+          // Защитный bound: <= 200 на source per request.
+          Math.min(size + page * size + 1, 200), 0);
+      for (DependentsQuery.DependentRow r : rows) {
+        all.add(new RecordDependent(
+            ref.sourceDict(),
+            ref.sourceDisplayName(),
+            ref.sourceField(),
+            ref.onClose(),
+            r.id(),
+            r.businessKey(),
+            r.validFrom(),
+            r.validTo(),
+            r.createdAt()));
+      }
+    }
+
+    // Sort all dependents by createdAt DESC, then page slice.
+    all.sort((a, b) -> b.createdAt().compareTo(a.createdAt()));
+    int from = Math.min(page * size, all.size());
+    int to = Math.min(from + size, all.size());
+    List<RecordDependent> slice = all.subList(from, to);
+
+    return new RecordDependentsPage(
+        targetDictName, targetBusinessKey, slice, summary, total, page, size);
+  }
+
+  // === DTOs ===
+
+  /** Schema-level reverse FK pointer: source dict → target dict через field. */
+  public record SchemaDependent(
+      String sourceDict,
+      String sourceDisplayName,
+      String sourceField,
+      String targetField,
+      OnCloseAction onClose,
+      long activeRecordsInSourceDict) {}
+
+  /** Record-level dependent — конкретная active record что ссылается. */
+  public record RecordDependent(
+      String sourceDict,
+      String sourceDisplayName,
+      String sourceField,
+      OnCloseAction onClose,
+      java.util.UUID id,
+      String businessKey,
+      OffsetDateTime validFrom,
+      OffsetDateTime validTo,
+      OffsetDateTime createdAt) {}
+
+  /** Per-source summary — сколько dependents в каждом source dict. */
+  public record PerSourceSummary(
+      String sourceDict,
+      String sourceDisplayName,
+      String sourceField,
+      OnCloseAction onClose,
+      long count) {}
+
+  /** Paginated page + per-source aggregates. */
+  public record RecordDependentsPage(
+      String targetDict,
+      String targetBusinessKey,
+      List<RecordDependent> items,
+      List<PerSourceSummary> perSource,
+      long total,
+      int page,
+      int size) {}
+}

ordinis-rest-api/src/main/java/cloud/nstart/terravault/ordinis/restapi/web/LineageController.java

@@ -0,0 +1,94 @@
+package cloud.nstart.terravault.ordinis.restapi.web;
+
+import cloud.nstart.terravault.ordinis.auth.ScopeContext;
+import cloud.nstart.terravault.ordinis.restapi.error.OrdinisException;
+import cloud.nstart.terravault.ordinis.restapi.service.reference.LineageIndexService;
+import org.springframework.web.bind.annotation.GetMapping;
+import org.springframework.web.bind.annotation.PathVariable;
+import org.springframework.web.bind.annotation.RequestMapping;
+import org.springframework.web.bind.annotation.RequestParam;
+import org.springframework.web.bind.annotation.RestController;
+
+import java.util.List;
+
+/**
+ * Lineage / dependents read-side. Phase 1 dict-relationships-v2 epic.
+ *
+ * <p>Два endpoint'а:
+ * <ul>
+ *   <li>{@code GET /api/v1/dictionaries/{name}/dependents} — schema-level
+ *       reverse FK list ("какие словари ссылаются на этот словарь и через
+ *       какие поля").</li>
+ *   <li>{@code GET /api/v1/records/{dict}/{businessKey}/dependents} —
+ *       record-level dependents (paginated, per-source counts).</li>
+ * </ul>
+ *
+ * <p>Errors:
+ * <ul>
+ *   <li>404 {@code dictionary_not_found} — target dict не существует или
+ *       не виден caller'у в его scopes (одинаковая ошибка чтобы не leak'ало
+ *       existence).</li>
+ *   <li>404 {@code record_not_active} — нет active record на now().</li>
+ *   <li>400 {@code invalid_pagination} — page/size вне допустимого диапазона.</li>
+ * </ul>
+ */
+@RestController
+@RequestMapping("/api/v1")
+public class LineageController {
+
+  private final LineageIndexService lineage;
+  private final ScopeContext scopeContext;
+
+  public LineageController(LineageIndexService lineage, ScopeContext scopeContext) {
+    this.lineage = lineage;
+    this.scopeContext = scopeContext;
+  }
+
+  /**
+   * Schema-level reverse FK list. Empty list если target не существует
+   * или не виден — consistent с CRUD endpoints (no leak existence).
+   *
+   * <p>Возврат: list объектов
+   * {@code {sourceDict, sourceDisplayName, sourceField, targetField,
+   *  onClose, activeRecordsInSourceDict}}.
+   */
+  @GetMapping("/dictionaries/{name}/dependents")
+  public List<LineageIndexService.SchemaDependent> schemaDependents(
+      @PathVariable String name) {
+    return lineage.findSchemaDependents(name, scopeContext.currentScopes());
+  }
+
+  /**
+   * Record-level dependents для конкретной record (target_dict + business_key).
+   *
+   * <p>Pagination: {@code page} (default 0), {@code size} (default 50, max 200).
+   *
+   * @return объект {@code {targetDict, targetBusinessKey, items[], perSource[],
+   *  total, page, size}}.
+   */
+  @GetMapping("/records/{dict}/{businessKey}/dependents")
+  public LineageIndexService.RecordDependentsPage recordDependents(
+      @PathVariable String dict,
+      @PathVariable String businessKey,
+      @RequestParam(defaultValue = "0") int page,
+      @RequestParam(defaultValue = "50") int size) {
+    if (page < 0 || size < 1 || size > 200) {
+      throw OrdinisException.badRequest(
+          "invalid_pagination",
+          "page must be >= 0 and size 1..200; got page=" + page + " size=" + size);
+    }
+    try {
+      return lineage.findRecordDependents(
+          dict, businessKey, scopeContext.currentScopes(), page, size);
+    } catch (IllegalArgumentException e) {
+      String msg = e.getMessage() == null ? "" : e.getMessage();
+      if (msg.startsWith("dictionary not found")) {
+        throw OrdinisException.notFound("dictionary_not_found", msg);
+      }
+      if (msg.startsWith("active record not found")) {
+        throw OrdinisException.notFound("record_not_active", msg);
+      }
+      throw OrdinisException.badRequest("invalid_request", msg);
+    }
+  }
+}

ordinis-rest-api/src/test/java/cloud/nstart/terravault/ordinis/restapi/service/reference/LineageIndexServiceTest.java

@@ -0,0 +1,368 @@
+package cloud.nstart.terravault.ordinis.restapi.service.reference;
+
+import cloud.nstart.terravault.ordinis.domain.DataScope;
+import cloud.nstart.terravault.ordinis.domain.definition.DictionaryDefinition;
+import cloud.nstart.terravault.ordinis.domain.definition.DictionaryDefinitionRepository;
+import cloud.nstart.terravault.ordinis.domain.record.DependentsQuery;
+import cloud.nstart.terravault.ordinis.domain.record.DictionaryRecord;
+import cloud.nstart.terravault.ordinis.domain.record.DictionaryRecordRepository;
+import com.fasterxml.jackson.databind.JsonNode;
+import com.fasterxml.jackson.databind.ObjectMapper;
+import org.junit.jupiter.api.BeforeEach;
+import org.junit.jupiter.api.Test;
+
+import java.time.OffsetDateTime;
+import java.util.List;
+import java.util.Optional;
+import java.util.Set;
+import java.util.UUID;
+
+import static org.assertj.core.api.Assertions.assertThat;
+import static org.assertj.core.api.Assertions.assertThatThrownBy;
+import static org.mockito.ArgumentMatchers.any;
+import static org.mockito.ArgumentMatchers.anyInt;
+import static org.mockito.ArgumentMatchers.anyString;
+import static org.mockito.ArgumentMatchers.eq;
+import static org.mockito.Mockito.mock;
+import static org.mockito.Mockito.when;
+
+/**
+ * Unit tests {@link LineageIndexService}. Покрывает:
+ *
+ * <ul>
+ *   <li>Schema-level reverse FK: not_found, scope hidden, malformed,
+ *       self-reference skip, sorting, onClose из x-references-on-close.</li>
+ *   <li>Record-level: not_active record → exception, no schema refs → empty,
+ *       perSource counts + paginated items, pagination across multiple
+ *       sources.</li>
+ * </ul>
+ *
+ * <p>Mockito не может mock'нуть DictionaryDefinition / DictionaryRecord
+ * на JDK 25 (subclass mock-maker incompat). Используем real entities.
+ */
+class LineageIndexServiceTest {
+
+  private static final ObjectMapper OM = new ObjectMapper();
+
+  private DictionaryDefinitionRepository defRepo;
+  private DictionaryRecordRepository recRepo;
+  private DependentsQuery depsQuery;
+  private LineageIndexService service;
+
+  @BeforeEach
+  void setUp() {
+    defRepo = mock(DictionaryDefinitionRepository.class);
+    recRepo = mock(DictionaryRecordRepository.class);
+    depsQuery = mock(DependentsQuery.class);
+    service = new LineageIndexService(defRepo, recRepo, depsQuery);
+  }
+
+  // === Schema-level findSchemaDependents ===
+
+  @Test
+  void schemaDeps_targetNotFound_returnsEmpty() {
+    when(defRepo.findByName("missing")).thenReturn(Optional.empty());
+    var result = service.findSchemaDependents("missing", Set.of(DataScope.PUBLIC));
+    assertThat(result).isEmpty();
+  }
+
+  @Test
+  void schemaDeps_targetScopeHidden_returnsEmpty() {
+    var target = mockDef("internal", DataScope.INTERNAL);
+    when(defRepo.findByName("internal")).thenReturn(Optional.of(target));
+    var result = service.findSchemaDependents("internal", Set.of(DataScope.PUBLIC));
+    assertThat(result).isEmpty();
+  }
+
+  @Test
+  void schemaDeps_oneReferenceFound_returnsIt() throws Exception {
+    var target = mockDef("satellite_type", DataScope.PUBLIC);
+    var source = mockDefWithSchema("spacecraft", DataScope.PUBLIC, """
+        {"properties":{"satTypeCode":{"type":"string","x-references":"satellite_type.code"}}}""");
+    when(defRepo.findByName("satellite_type")).thenReturn(Optional.of(target));
+    when(defRepo.findAll()).thenReturn(List.of(target, source));
+    List<Object[]> rows = new java.util.ArrayList<>();
+    rows.add(new Object[]{source.getId(), 7L});
+    when(recRepo.countActiveGroupedByDictionary(any(), any())).thenReturn(rows);
+
+    var result = service.findSchemaDependents("satellite_type", Set.of(DataScope.PUBLIC));
+
+    assertThat(result).hasSize(1);
+    var dep = result.get(0);
+    assertThat(dep.sourceDict()).isEqualTo("spacecraft");
+    assertThat(dep.sourceField()).isEqualTo("satTypeCode");
+    assertThat(dep.targetField()).isEqualTo("code");
+    assertThat(dep.onClose()).isEqualTo(OnCloseAction.BLOCK); // default
+    assertThat(dep.activeRecordsInSourceDict()).isEqualTo(7L);
+  }
+
+  @Test
+  void schemaDeps_onCloseFromSchema_propagated() throws Exception {
+    var target = mockDef("ground_station", DataScope.PUBLIC);
+    var source = mockDefWithSchema("contact_session", DataScope.PUBLIC, """
+        {"properties":{
+          "stationCode":{"type":"string",
+            "x-references":"ground_station.code",
+            "x-references-on-close":"cascade"}
+        }}""");
+    when(defRepo.findByName("ground_station")).thenReturn(Optional.of(target));
+    when(defRepo.findAll()).thenReturn(List.of(target, source));
+    when(recRepo.countActiveGroupedByDictionary(any(), any())).thenReturn(List.of());
+
+    var result = service.findSchemaDependents("ground_station", Set.of(DataScope.PUBLIC));
+
+    assertThat(result).hasSize(1);
+    assertThat(result.get(0).onClose()).isEqualTo(OnCloseAction.CASCADE);
+  }
+
+  @Test
+  void schemaDeps_multipleSourcesSorted() throws Exception {
+    var target = mockDef("dict_t", DataScope.PUBLIC);
+    var srcA = mockDefWithSchema("zeta_dict", DataScope.PUBLIC,
+        "{\"properties\":{\"r\":{\"x-references\":\"dict_t.f\"}}}");
+    var srcB = mockDefWithSchema("alpha_dict", DataScope.PUBLIC,
+        "{\"properties\":{\"r\":{\"x-references\":\"dict_t.f\"}}}");
+    when(defRepo.findByName("dict_t")).thenReturn(Optional.of(target));
+    when(defRepo.findAll()).thenReturn(List.of(target, srcA, srcB));
+    when(recRepo.countActiveGroupedByDictionary(any(), any())).thenReturn(List.of());
+
+    var result = service.findSchemaDependents("dict_t", Set.of(DataScope.PUBLIC));
+
+    assertThat(result).extracting(LineageIndexService.SchemaDependent::sourceDict)
+        .containsExactly("alpha_dict", "zeta_dict");
+  }
+
+  @Test
+  void schemaDeps_sourceScopeHidden_skipped() throws Exception {
+    var target = mockDef("dict_t", DataScope.PUBLIC);
+    var visibleSrc = mockDefWithSchema("public_src", DataScope.PUBLIC,
+        "{\"properties\":{\"r\":{\"x-references\":\"dict_t.f\"}}}");
+    var hiddenSrc = mockDefWithSchema("internal_src", DataScope.INTERNAL,
+        "{\"properties\":{\"r\":{\"x-references\":\"dict_t.f\"}}}");
+    when(defRepo.findByName("dict_t")).thenReturn(Optional.of(target));
+    when(defRepo.findAll()).thenReturn(List.of(target, visibleSrc, hiddenSrc));
+    when(recRepo.countActiveGroupedByDictionary(any(), any())).thenReturn(List.of());
+
+    var result = service.findSchemaDependents("dict_t", Set.of(DataScope.PUBLIC));
+
+    assertThat(result).extracting(LineageIndexService.SchemaDependent::sourceDict)
+        .containsExactly("public_src");
+  }
+
+  @Test
+  void schemaDeps_selfReference_skipped() throws Exception {
+    // Schema указывает на свой же dict — не должно зацикливать.
+    var selfRef = mockDefWithSchema("hierarchy", DataScope.PUBLIC,
+        "{\"properties\":{\"parentCode\":{\"x-references\":\"hierarchy.code\"}}}");
+    when(defRepo.findByName("hierarchy")).thenReturn(Optional.of(selfRef));
+    when(defRepo.findAll()).thenReturn(List.of(selfRef));
+    when(recRepo.countActiveGroupedByDictionary(any(), any())).thenReturn(List.of());
+
+    var result = service.findSchemaDependents("hierarchy", Set.of(DataScope.PUBLIC));
+
+    assertThat(result).isEmpty();
+  }
+
+  @Test
+  void schemaDeps_malformedRef_skippedSilently() throws Exception {
+    var target = mockDef("dict_t", DataScope.PUBLIC);
+    var srcGood = mockDefWithSchema("good_src", DataScope.PUBLIC,
+        "{\"properties\":{\"r\":{\"x-references\":\"dict_t.f\"}}}");
+    var srcBad = mockDefWithSchema("bad_src", DataScope.PUBLIC,
+        "{\"properties\":{\"r\":{\"x-references\":\"malformed-no-dot\"}}}");
+    when(defRepo.findByName("dict_t")).thenReturn(Optional.of(target));
+    when(defRepo.findAll()).thenReturn(List.of(target, srcGood, srcBad));
+    when(recRepo.countActiveGroupedByDictionary(any(), any())).thenReturn(List.of());
+
+    var result = service.findSchemaDependents("dict_t", Set.of(DataScope.PUBLIC));
+
+    assertThat(result).extracting(LineageIndexService.SchemaDependent::sourceDict)
+        .containsExactly("good_src");
+  }
+
+  @Test
+  void schemaDeps_targetSchemaWithoutProperties_skipped() throws Exception {
+    var target = mockDef("dict_t", DataScope.PUBLIC);
+    var src = mockDefWithSchema("src", DataScope.PUBLIC, "{\"type\":\"object\"}"); // no properties
+    when(defRepo.findByName("dict_t")).thenReturn(Optional.of(target));
+    when(defRepo.findAll()).thenReturn(List.of(target, src));
+    when(recRepo.countActiveGroupedByDictionary(any(), any())).thenReturn(List.of());
+
+    var result = service.findSchemaDependents("dict_t", Set.of(DataScope.PUBLIC));
+
+    assertThat(result).isEmpty();
+  }
+
+  // === Record-level findRecordDependents ===
+
+  @Test
+  void recordDeps_invalidPagination_throws() {
+    assertThatThrownBy(() -> service.findRecordDependents(
+        "d", "k", Set.of(DataScope.PUBLIC), -1, 50))
+        .isInstanceOf(IllegalArgumentException.class)
+        .hasMessageContaining("page");
+    assertThatThrownBy(() -> service.findRecordDependents(
+        "d", "k", Set.of(DataScope.PUBLIC), 0, 0))
+        .isInstanceOf(IllegalArgumentException.class)
+        .hasMessageContaining("size");
+    assertThatThrownBy(() -> service.findRecordDependents(
+        "d", "k", Set.of(DataScope.PUBLIC), 0, 201))
+        .isInstanceOf(IllegalArgumentException.class)
+        .hasMessageContaining("size");
+  }
+
+  @Test
+  void recordDeps_dictNotFound_throws() {
+    when(defRepo.findByName("missing")).thenReturn(Optional.empty());
+    assertThatThrownBy(() -> service.findRecordDependents(
+        "missing", "k", Set.of(DataScope.PUBLIC), 0, 50))
+        .isInstanceOf(IllegalArgumentException.class)
+        .hasMessageContaining("dictionary not found");
+  }
+
+  @Test
+  void recordDeps_dictScopeHidden_throwsAsNotFound() {
+    var target = mockDef("internal", DataScope.INTERNAL);
+    when(defRepo.findByName("internal")).thenReturn(Optional.of(target));
+    assertThatThrownBy(() -> service.findRecordDependents(
+        "internal", "k", Set.of(DataScope.PUBLIC), 0, 50))
+        .isInstanceOf(IllegalArgumentException.class)
+        .hasMessageContaining("dictionary not found");
+  }
+
+  @Test
+  void recordDeps_recordNotActive_throws() {
+    var target = mockDef("dict_t", DataScope.PUBLIC);
+    when(defRepo.findByName("dict_t")).thenReturn(Optional.of(target));
+    when(recRepo.findActiveAt(eq(target.getId()), eq("missing-key"), any()))
+        .thenReturn(Optional.empty());
+
+    assertThatThrownBy(() -> service.findRecordDependents(
+        "dict_t", "missing-key", Set.of(DataScope.PUBLIC), 0, 50))
+        .isInstanceOf(IllegalArgumentException.class)
+        .hasMessageContaining("active record not found");
+  }
+
+  @Test
+  void recordDeps_noSchemaRefs_returnsEmptyPage() throws Exception {
+    var target = mockDef("dict_t", DataScope.PUBLIC);
+    when(defRepo.findByName("dict_t")).thenReturn(Optional.of(target));
+    when(defRepo.findAll()).thenReturn(List.of(target));
+    when(recRepo.countActiveGroupedByDictionary(any(), any())).thenReturn(List.of());
+
+    JsonNode data = OM.readTree("{\"code\":\"X-001\"}");
+    DictionaryRecord rec = mockRecord(target.getId(), "X-001", data);
+    when(recRepo.findActiveAt(eq(target.getId()), eq("X-001"), any()))
+        .thenReturn(Optional.of(rec));
+
+    var page = service.findRecordDependents(
+        "dict_t", "X-001", Set.of(DataScope.PUBLIC), 0, 50);
+
+    assertThat(page.items()).isEmpty();
+    assertThat(page.perSource()).isEmpty();
+    assertThat(page.total()).isZero();
+  }
+
+  @Test
+  void recordDeps_oneSourceWithDependents_returnsItemsAndPerSource() throws Exception {
+    var target = mockDef("satellite_type", DataScope.PUBLIC);
+    var source = mockDefWithSchema("spacecraft", DataScope.PUBLIC, """
+        {"properties":{"satTypeCode":{"type":"string","x-references":"satellite_type.code"}}}""");
+    when(defRepo.findByName("satellite_type")).thenReturn(Optional.of(target));
+    when(defRepo.findByName("spacecraft")).thenReturn(Optional.of(source));
+    when(defRepo.findAll()).thenReturn(List.of(target, source));
+    when(recRepo.countActiveGroupedByDictionary(any(), any())).thenReturn(List.of());
+
+    JsonNode targetData = OM.readTree("{\"code\":\"SAR-X\"}");
+    DictionaryRecord targetRec = mockRecord(target.getId(), "SAR-X", targetData);
+    when(recRepo.findActiveAt(eq(target.getId()), eq("SAR-X"), any()))
+        .thenReturn(Optional.of(targetRec));
+
+    when(depsQuery.countActive(eq(source.getId()), eq("satTypeCode"), eq("SAR-X"), any()))
+        .thenReturn(2L);
+    OffsetDateTime t1 = OffsetDateTime.parse("2026-01-01T00:00:00Z");
+    OffsetDateTime t2 = OffsetDateTime.parse("2026-02-01T00:00:00Z");
+    when(depsQuery.findActive(eq(source.getId()), eq("satTypeCode"), eq("SAR-X"),
+        any(), anyInt(), anyInt())).thenReturn(List.of(
+            new DependentsQuery.DependentRow(
+                UUID.randomUUID(), source.getId(), "SC-001", t1, OffsetDateTime.MAX, t2),
+            new DependentsQuery.DependentRow(
+                UUID.randomUUID(), source.getId(), "SC-002", t1, OffsetDateTime.MAX, t1)));
+
+    var page = service.findRecordDependents(
+        "satellite_type", "SAR-X", Set.of(DataScope.PUBLIC), 0, 50);
+
+    assertThat(page.total()).isEqualTo(2L);
+    assertThat(page.perSource()).hasSize(1);
+    assertThat(page.perSource().get(0).count()).isEqualTo(2L);
+    assertThat(page.perSource().get(0).sourceDict()).isEqualTo("spacecraft");
+    assertThat(page.items()).hasSize(2);
+    // sorted by createdAt DESC: SC-001 (t2) first
+    assertThat(page.items().get(0).businessKey()).isEqualTo("SC-001");
+    assertThat(page.items().get(1).businessKey()).isEqualTo("SC-002");
+  }
+
+  @Test
+  void recordDeps_pagination_slicesCorrectly() throws Exception {
+    var target = mockDef("dict_t", DataScope.PUBLIC);
+    var source = mockDefWithSchema("src", DataScope.PUBLIC,
+        "{\"properties\":{\"ref\":{\"x-references\":\"dict_t.code\"}}}");
+    when(defRepo.findByName("dict_t")).thenReturn(Optional.of(target));
+    when(defRepo.findByName("src")).thenReturn(Optional.of(source));
+    when(defRepo.findAll()).thenReturn(List.of(target, source));
+    when(recRepo.countActiveGroupedByDictionary(any(), any())).thenReturn(List.of());
+
+    JsonNode targetData = OM.readTree("{\"code\":\"K\"}");
+    when(recRepo.findActiveAt(eq(target.getId()), eq("K"), any()))
+        .thenReturn(Optional.of(mockRecord(target.getId(), "K", targetData)));
+
+    when(depsQuery.countActive(any(), anyString(), anyString(), any())).thenReturn(5L);
+    // 5 rows, sorted later by createdAt DESC.
+    var rows = new java.util.ArrayList<DependentsQuery.DependentRow>();
+    for (int i = 0; i < 5; i++) {
+      rows.add(new DependentsQuery.DependentRow(
+          UUID.randomUUID(), source.getId(), "K-" + i,
+          OffsetDateTime.parse("2026-01-01T00:00:00Z"), OffsetDateTime.MAX,
+          OffsetDateTime.parse("2026-01-0" + (i + 1) + "T00:00:00Z")));
+    }
+    when(depsQuery.findActive(any(), anyString(), anyString(), any(), anyInt(), anyInt()))
+        .thenReturn(rows);
+
+    var page1 = service.findRecordDependents("dict_t", "K", Set.of(DataScope.PUBLIC), 0, 2);
+    assertThat(page1.items()).hasSize(2);
+    assertThat(page1.items().get(0).businessKey()).isEqualTo("K-4"); // newest
+    assertThat(page1.total()).isEqualTo(5L);
+
+    var page2 = service.findRecordDependents("dict_t", "K", Set.of(DataScope.PUBLIC), 1, 2);
+    assertThat(page2.items()).hasSize(2);
+    assertThat(page2.items().get(0).businessKey()).isEqualTo("K-2");
+
+    var page3 = service.findRecordDependents("dict_t", "K", Set.of(DataScope.PUBLIC), 2, 2);
+    assertThat(page3.items()).hasSize(1);
+    assertThat(page3.items().get(0).businessKey()).isEqualTo("K-0"); // oldest
+  }
+
+  // === Helpers ===
+
+  private static DictionaryDefinition mockDef(String name, DataScope scope) {
+    return mockDefWithSchema(name, scope, "{\"type\":\"object\"}");
+  }
+
+  private static DictionaryDefinition mockDefWithSchema(String name, DataScope scope, String schemaJson) {
+    JsonNode schema;
+    try {
+      schema = OM.readTree(schemaJson);
+    } catch (Exception e) {
+      throw new RuntimeException(e);
+    }
+    var def = new DictionaryDefinition(UUID.randomUUID(), name, scope, schema);
+    def.setDisplayName(name);
+    return def;
+  }
+
+  private static DictionaryRecord mockRecord(UUID dictId, String businessKey, JsonNode data) {
+    return new DictionaryRecord(
+        UUID.randomUUID(), dictId, businessKey, data, DataScope.PUBLIC,
+        OffsetDateTime.parse("2026-01-01T00:00:00Z"));
+  }
+}

ordinis-rest-api/src/test/java/cloud/nstart/terravault/ordinis/restapi/service/reference/ReferenceCloseRegressionTest.java

@@ -0,0 +1,157 @@
+package cloud.nstart.terravault.ordinis.restapi.service.reference;
+
+import cloud.nstart.terravault.ordinis.domain.DataScope;
+import cloud.nstart.terravault.ordinis.domain.definition.DictionaryDefinition;
+import cloud.nstart.terravault.ordinis.domain.definition.DictionaryDefinitionRepository;
+import cloud.nstart.terravault.ordinis.domain.record.DictionaryRecord;
+import cloud.nstart.terravault.ordinis.domain.record.DictionaryRecordRepository;
+import cloud.nstart.terravault.ordinis.validation.ValidationError;
+import com.fasterxml.jackson.databind.JsonNode;
+import com.fasterxml.jackson.databind.ObjectMapper;
+import org.junit.jupiter.api.BeforeEach;
+import org.junit.jupiter.api.Test;
+
+import java.time.OffsetDateTime;
+import java.util.List;
+import java.util.Optional;
+import java.util.UUID;
+
+import static org.assertj.core.api.Assertions.assertThat;
+import static org.mockito.ArgumentMatchers.any;
+import static org.mockito.ArgumentMatchers.anyString;
+import static org.mockito.ArgumentMatchers.eq;
+import static org.mockito.Mockito.mock;
+import static org.mockito.Mockito.when;
+
+/**
+ * REGRESSION mandatory: existing close path с null {@code x-references-on-close}
+ * field MUST behave identical к v1.2.1.
+ *
+ * <p>Покрывает gap: bundle v1.2.1 (без новой schema field) loaded в новый build —
+ * убедиться что:
+ * <ol>
+ *   <li>{@link ReferenceValidator#validate} возвращает те же error codes что и
+ *       до Phase 1 prep commit (никаких новых x-references-on-close-related
+ *       codes не появилось).</li>
+ *   <li>{@link OnCloseAction#BLOCK} default'ится автоматически — schema parser
+ *       не падает и не меняет error response shape.</li>
+ *   <li>Active record на которое есть references просто закрывается — на этом
+ *       этапе close не обращается к onClose (cascade engine = Phase 3).</li>
+ * </ol>
+ *
+ * <p>Без этого теста риск молча сломать live consumers (см. design doc test
+ * plan REGRESSION section).
+ */
+class ReferenceCloseRegressionTest {
+
+  private static final ObjectMapper OM = new ObjectMapper();
+
+  private DictionaryDefinitionRepository defRepo;
+  private DictionaryRecordRepository recRepo;
+  private ReferenceValidator validator;
+
+  @BeforeEach
+  void setUp() {
+    defRepo = mock(DictionaryDefinitionRepository.class);
+    recRepo = mock(DictionaryRecordRepository.class);
+    validator = new ReferenceValidator(defRepo, recRepo);
+  }
+
+  /**
+   * v1.2.1 schema = x-references есть, x-references-on-close нет.
+   * validate() должен:
+   * — пропустить запись с валидным reference (no errors);
+   * — отклонить запись с invalid reference с тем же кодом
+   *   {@code x_references_target_not_found} (без новых fields в error).
+   */
+  @Test
+  void v1_2_1_schemaWithoutOnClose_validateBehavesIdentical() throws Exception {
+    UUID targetId = UUID.randomUUID();
+    DictionaryDefinition target = new DictionaryDefinition(
+        targetId, "satellite_type", DataScope.PUBLIC, OM.readTree("{}"));
+    when(defRepo.findByName("satellite_type")).thenReturn(Optional.of(target));
+    // Active record с code=LEO-001 существует, NONEXISTENT — нет.
+    when(recRepo.findActiveByJsonField(eq(targetId), any(), eq("code"), eq("LEO-001")))
+        .thenReturn(List.of(stubRecord(targetId)));
+    when(recRepo.findActiveByJsonField(any(), any(), anyString(), eq("NONEXISTENT")))
+        .thenReturn(List.of());
+
+    // V1.2.1 schema: x-references есть, x-references-on-close ОТСУТСТВУЕТ.
+    JsonNode schema = OM.readTree("""
+        {"type":"object","properties":{
+          "satTypeCode":{"type":"string","x-references":"satellite_type.code"}
+        }}""");
+
+    // Valid reference — нет errors.
+    JsonNode goodData = OM.readTree("{\"satTypeCode\":\"LEO-001\"}");
+    List<ValidationError> goodErrors = validator.validate(
+        schema, goodData, List.of(DataScope.PUBLIC), OffsetDateTime.now());
+    assertThat(goodErrors).isEmpty();
+
+    // Invalid reference — точно такой же error code и message structure как в v1.
+    JsonNode badData = OM.readTree("{\"satTypeCode\":\"NONEXISTENT\"}");
+    List<ValidationError> badErrors = validator.validate(
+        schema, badData, List.of(DataScope.PUBLIC), OffsetDateTime.now());
+    assertThat(badErrors).hasSize(1);
+    ValidationError err = badErrors.get(0);
+    assertThat(err.code()).isEqualTo("x_references_target_not_found");
+    assertThat(err.path()).isEqualTo("/satTypeCode");
+    assertThat(err.message()).contains("satellite_type");
+    assertThat(err.message()).contains("NONEXISTENT");
+    // Никаких новых полей в error: x_references_on_close_* кодов не появилось.
+    assertThat(err.code()).doesNotContain("on_close");
+    assertThat(err.code()).doesNotContain("cascade");
+  }
+
+  /**
+   * Schema parser применяется к v1.2.1 bundle — onClose default'ится в BLOCK
+   * автоматически. Никаких малформ-related parse errors.
+   */
+  @Test
+  void v1_2_1_schemaParser_defaultsOnCloseToBlock() throws Exception {
+    JsonNode propV1 = OM.readTree("""
+        {"type":"string","x-references":"d.f"}""");
+    var parsed = ReferenceValidator.parseRef("d.f", propV1);
+    assertThat(parsed.onClose()).isEqualTo(OnCloseAction.BLOCK);
+    assertThat(parsed.dictionaryName()).isEqualTo("d");
+    assertThat(parsed.fieldName()).isEqualTo("f");
+  }
+
+  /**
+   * Defensive: legacy single-arg overload {@code parseRef(spec)} (без property
+   * schema) сохраняет старый contract. Используется внешними тестами и orphan
+   * scanner (не v1 API но shared parser).
+   */
+  @Test
+  void v1_2_1_schemaParser_singleArgStillWorks() {
+    var parsed = ReferenceValidator.parseRef("satellite_type.code");
+    assertThat(parsed.onClose()).isEqualTo(OnCloseAction.BLOCK);
+    assertThat(parsed.dictionaryName()).isEqualTo("satellite_type");
+    assertThat(parsed.fieldName()).isEqualTo("code");
+  }
+
+  /**
+   * Schema без x-references вообще — full no-op, validate возвращает пустой
+   * list errors. Покрывает 30+ legacy справочников что не имеют FK.
+   */
+  @Test
+  void v1_2_1_schemaWithoutXReferences_isFullNoOp() throws Exception {
+    JsonNode schema = OM.readTree("""
+        {"type":"object","properties":{
+          "name":{"type":"string"},"code":{"type":"string"}
+        }}""");
+    JsonNode data = OM.readTree("{\"name\":\"Spacecraft Alpha\",\"code\":\"SC-001\"}");
+
+    List<ValidationError> errors = validator.validate(
+        schema, data, List.of(DataScope.PUBLIC), OffsetDateTime.now());
+    assertThat(errors).isEmpty();
+  }
+
+  // === Helpers ===
+
+  private static DictionaryRecord stubRecord(UUID dictId) {
+    return new DictionaryRecord(
+        UUID.randomUUID(), dictId, "stub-key",
+        OM.createObjectNode(), DataScope.PUBLIC, OffsetDateTime.now());
+  }
+}