diff --git a/docs/ingestion.md b/docs/ingestion.md
index b1575f6..afaae5d 100644
--- a/docs/ingestion.md
+++ b/docs/ingestion.md
@@ -79,9 +79,8 @@ units, multivalued, required, pattern}}}`.
 Converts the dict into real, hash-identified objects:
 
 - **Properties are built first.** Each one's `hash_id` is computed from
-  `name`/`description`/`range`/`units` only — `slot_uri` and
-  `registry_version` are stored on the object but excluded from the hash
-  (they're provenance/operational metadata, not identity).
+  `name`/`description`/`range`/`units` only — `slot_uri` is stored on the
+  object but excluded from the hash (it's origin metadata, not identity).
 - **Classes reference their properties by hash_id**, sorted. This is why a
   class's own hash depends on its full induced property set — two classes
   with the same properties (regardless of declaration order) hash the same.
@@ -91,8 +90,8 @@ Converts the dict into real, hash-identified objects:
   have every `is_a` target resolvable within its own import closure (that's
   `SchemaView`'s job), so this recursion always terminates cleanly.
 - **Every entity gets one `ProvenanceEntry`** for this ingestion — `source`,
-  `attributed_to` (agent), `generated_at`, `activity` — entirely separate
-  from the hash computation.
+  `attributed_to` (agent), `generated_at`, `activity`, `registry_version` —
+  entirely separate from the hash computation.
 
 ### 3. `write_registry_entities()` + `write_structural_edges()` (`neuro_ghost/db.py`)
 
@@ -116,9 +115,8 @@ and `seed.py` — schema.org is ingested through the exact same path, just with
 | `range`, `units` | `RegistryProperty` | Identity-defining. |
 | `properties`, `is_a`, `mixins` | `RegistryClass` | Identity-defining — all stored as hash_id references. |
 | `class_uri` / `slot_uri` | `RegistryClass` / `RegistryProperty` | Ontology IRI preserved from the source. **Not** part of the hash — two schemas using different IRIs (or none) for the same content still collapse to one node. |
-| `registry_version` | both | Operational stamp. Not part of the hash. |
 | `provenance` | both | List of `ProvenanceEntry`. Accumulates, never affects `hash_id`. |
-| `source`, `attributed_to`, `generated_at`, `activity`, `derived_from` | `ProvenanceEntry` | PROV-O–grounded (`slot_uri: prov:wasAttributedTo` etc.) |
+| `source`, `attributed_to`, `generated_at`, `activity`, `derived_from`, `registry_version` | `ProvenanceEntry` | PROV-O–grounded (`slot_uri: prov:wasAttributedTo` etc.). `registry_version` lives here, not on the entity — the same entity can be attested by different sources at different times, each under a different registry version, so a single scalar on the entity doesn't fit (same reasoning as dropping `source_label`). |
 
 Field names were deliberately aligned with LinkML's own metamodel
 (`description`, `range`, `class_uri`/`slot_uri`, `is_a`, `abstract`) rather
@@ -167,6 +165,33 @@ Saying "parse_linkml extracts X" and "the registry stores X" are different
 claims — test each layer for what it actually is, not for what you assume
 the other layer does with it.
 
+`tests/test_ingest_registry.py::test_required_does_not_affect_property_identity`
+takes this one step further, end to end: two schemas declare the exact same
+`age` slot except one marks it `required: true`. Ingesting both must produce
+exactly one `RegistryProperty` node, not two — proving `required` doesn't
+leak into identity, in the real graph, not just in an isolated object.
+
+## Open question: when should ProvenanceEntry.registry_version be set?
+
+The whole-registry version (the semver in `data/registry.json`, e.g. `1.7.0`)
+only gets computed once, at the very end of a submission, when
+`export_json.py` reads the last entry in `data/provenance.json` and bumps it.
+`seed.py`/`ingest_linkml.py`/`align.py` all run *before* that — so the version
+an entity is actually going to belong to doesn't exist yet at the moment it's
+ingested.
+
+That rules out the obvious fix ("pass the current, pre-bump version to
+ingest_linkml.py/align.py") — it would record the version this submission is
+replacing, not the one it's part of. The consistent fix would be computing
+the bump once, up front (alongside where the Action already determines
+`bump` type in "Parse issue metadata"), and threading that single value
+through every step including `export_json.py` — instead of `export_json.py`
+computing it independently after the fact.
+
+**Not implemented.** `schema_submission.yml` does not pass `--registry-version`
+to `ingest_linkml.py` or `align.py` — those entities' `ProvenanceEntry.
+registry_version` stays `None` for now, pending a decision on the above.
+
 ## Known gaps (as of this writing)
 
 - **`index.html`** (frontend) still expects the pre-rework JSON shape
diff --git a/neuro_ghost/export_json.py b/neuro_ghost/export_json.py
index 48261cb..a539136 100644
--- a/neuro_ghost/export_json.py
+++ b/neuro_ghost/export_json.py
@@ -96,8 +96,8 @@ def export_snapshot(conn, registry_version: str) -> dict:
             "hash_id":          hash_id,
             "iri":              class_uri or "",
             "name":             name or "",
-            "definition":  desc or "",
-            "is_abstract": bool(is_abstract),
+            "definition":       desc or "",
+            "is_abstract":      bool(is_abstract),
             "sources":          _attesting_sources(conn, "RegistryClass", "HAS_PROVENANCE", hash_id),
             "properties": [
                 {