diff --git a/harness/tests/test_qdrant_vector_store.py b/harness/tests/test_qdrant_vector_store.py
index af575b03..a4c6556a 100644
--- a/harness/tests/test_qdrant_vector_store.py
+++ b/harness/tests/test_qdrant_vector_store.py
@@ -892,3 +892,73 @@ def test_scroll_accepts_null_or_missing_terminal_offset(next_page_offset):
         expected_dimension=1024, request=request,
     )
     assert store.existing_hashes("memory", ["memory"]) == {}
+
+
+_QUERY_PAYLOAD = {
+    "record_key": "memory:1",
+    "record_kind": "memory",
+    "kind": "memory",
+    "ref": "ref:memory:1",
+    "title": "title:memory:1",
+    "content": "content:memory:1",
+}
+
+
+def _query_response_store(points):
+    fake = FakeQdrantHttp()
+    fake.collection = {"vectors": {"size": 1024, "distance": "Cosine"}}
+    fake.payload_indexes = set(_REQUIRED_INDEXES)
+    original = fake.request
+
+    def request(method, url, **kwargs):
+        if method == "POST" and url.endswith("/points/query"):
+            return FakeResponse(200, {"result": {"points": points}})
+        return original(method, url, **kwargs)
+
+    store = _store(fake)
+    store._request = request
+    return store
+
+
+@pytest.mark.parametrize("leaf", [
+    "record_key", "record_kind", "kind", "ref", "title", "content",
+])
+def test_query_rejects_non_string_required_vector_hit_payload_leaves(leaf):
+    payload = {**_QUERY_PAYLOAD, leaf: 1}
+    store = _query_response_store([{"id": "p1", "score": 0.9, "payload": payload}])
+
+    with pytest.raises(VectorResponseError):
+        store.search(["memory"], [0.2] * 1024, limit=1, kinds=["memory"])
+
+
+@pytest.mark.parametrize("leaf", [
+    "record_key", "record_kind", "kind", "ref", "title", "content",
+])
+def test_query_rejects_missing_required_vector_hit_payload_leaves(leaf):
+    payload = {key: value for key, value in _QUERY_PAYLOAD.items() if key != leaf}
+    store = _query_response_store([{"id": "p1", "score": 0.9, "payload": payload}])
+
+    with pytest.raises(VectorResponseError):
+        store.search(["memory"], [0.2] * 1024, limit=1, kinds=["memory"])
+
+
+def test_query_rejects_heterogeneous_record_keys_before_sorting_hits():
+    store = _query_response_store([
+        {"id": "p1", "score": 0.9, "payload": _QUERY_PAYLOAD},
+        {
+            "id": "p2",
+            "score": 0.8,
+            "payload": {**_QUERY_PAYLOAD, "record_key": 2},
+        },
+    ])
+
+    with pytest.raises(VectorResponseError):
+        store.search(["memory"], [0.2] * 1024, limit=2, kinds=["memory"])
+
+
+@pytest.mark.parametrize("score", [True, False, 10**1000, float("nan"), float("inf"), float("-inf")])
+def test_query_rejects_scores_that_cannot_be_finite_float(score):
+    store = _query_response_store([{"id": "p1", "score": score, "payload": _QUERY_PAYLOAD}])
+
+    with pytest.raises(VectorResponseError):
+        store.search(["memory"], [0.2] * 1024, limit=1, kinds=["memory"])
diff --git a/harness/tht/adapters/vector/qdrant.py b/harness/tht/adapters/vector/qdrant.py
index 91045b2f..c1f91e09 100644
--- a/harness/tht/adapters/vector/qdrant.py
+++ b/harness/tht/adapters/vector/qdrant.py
@@ -1,5 +1,6 @@
 """Qdrant-backed vector store for one workspace-owned semantic collection."""
 
+import math
 import re
 from collections.abc import Callable
 from uuid import NAMESPACE_URL, uuid5
@@ -477,16 +478,25 @@ class QdrantVectorStore:
             raise VectorResponseError("Qdrant returned malformed query response")
         payload = point.get("payload")
         score = point.get("score")
+        required_payload_leaves = (
+            "record_key", "record_kind", "kind", "ref", "title", "content"
+        )
         if (
             not isinstance(payload, dict)
             or type(score) not in (int, float)
             or any(
-                key in payload and not isinstance(payload[key], str)
-                for key in ("record_key", "record_kind", "kind", "ref", "title", "content")
+                key not in payload or not isinstance(payload[key], str)
+                for key in required_payload_leaves
             )
         ):
             raise VectorResponseError("Qdrant returned malformed query response")
-        return hit_from_metadata(float(score), payload)
+        try:
+            similarity = float(score)
+        except (OverflowError, ValueError) as exc:
+            raise VectorResponseError("Qdrant returned malformed query response") from exc
+        if not math.isfinite(similarity):
+            raise VectorResponseError("Qdrant returned malformed query response")
+        return hit_from_metadata(similarity, payload)
 
     def _call(self, method: str, path: str, payload: dict | None, allow_missing: bool = False) -> dict | None:
         try: