diff --git a/src/azul/service/manifest_service.py b/src/azul/service/manifest_service.py
index 1dd20de30..5d7489ce2 100644
--- a/src/azul/service/manifest_service.py
+++ b/src/azul/service/manifest_service.py
@@ -2011,6 +2011,28 @@ def qualify(qualifier, column_name, index=None):
             bundle_tsv_writer.writerow(row)
 
 
+@attrs.frozen(kw_only=True)
+class ReplicaKeys:
+    """
+    Most replicas contain a list of the replica's hubs. However, some entities
+    (e.g. projects) have too many hubs to track within the replica document.
+    Replicas of such entities are instead retrieved by their entity ID.
+    """
+    hub_id: str
+    implicit_hub_entity_id: str
+
+    @classmethod
+    def prepare_query(cls, keys: Iterable[Self]) -> Q:
+        terms = {'hub_ids': set(), 'entity_id': set()}
+        for self in keys:
+            terms['hub_ids'].add(self.hub_id)
+            terms['entity_id'].add(self.implicit_hub_entity_id)
+        return Q('bool', should=[
+            {'terms': {f'{field}.keyword': list(values)}}
+            for field, values in terms.items()
+        ])
+
+
 class VerbatimManifestGenerator(FileBasedManifestGenerator):
 
     @property
@@ -2043,17 +2065,13 @@ def included_fields(self) -> list[FieldPath]:
     def implicit_hub_type(self) -> str:
         return self.service.metadata_plugin(self.catalog).implicit_hub_type
 
-    def _replica_keys(self) -> Iterable[dict[str, str]]:
+    def _replica_keys(self) -> Iterable[ReplicaKeys]:
         request = self._create_request()
         for hit in request.scan():
-            yield {
-                # Most replicas track their hubs explicitly, however...
-                'hub_ids': hit['entity_id'],
-                # ... for projects and datasets, there are too many hubs to
-                # track them all in the replica, so they are instead retrieved
-                # by entity ID.
-                'entity_id': one(one(hit['contents'][self.implicit_hub_type])['document_id'])
-            }
+            yield ReplicaKeys(
+                hub_id=hit['entity_id'],
+                implicit_hub_entity_id=one(one(hit['contents'][self.implicit_hub_type])['document_id'])
+            )
 
     def _all_replicas(self) -> Iterable[JSON]:
         emitted_replica_ids = set()
@@ -2076,20 +2094,13 @@ def _all_replicas(self) -> Iterable[JSON]:
                         emitted_replica_ids.add(replica_id)
 
     def _join_replicas(self,
-                       keys_page: Iterable[dict[str, str]]
+                       keys_page: Iterable[ReplicaKeys]
                        ) -> Iterable[Hit]:
-        terms_by_field = defaultdict(set)
-        for keys in keys_page:
-            for field, term in keys.items():
-                terms_by_field[field].add(term)
+
         request = self.service.create_request(catalog=self.catalog,
                                               entity_type='replica',
                                               doc_type=DocumentType.replica)
-        request = request.query(Q('bool', should=[
-            {'terms': {f'{field}.keyword': list(terms)}}
-            for field, terms in terms_by_field.items()
-        ]
-                                  ))
+        request = request.query(ReplicaKeys.prepare_query(keys_page))
         return request.scan()
 
     def create_file(self) -> tuple[str, Optional[str]]: