diff --git a/src/calibre/ebooks/metadata/sources/identify.py b/src/calibre/ebooks/metadata/sources/identify.py
index 303bb2db6e..81f3e8247f 100644
--- a/src/calibre/ebooks/metadata/sources/identify.py
+++ b/src/calibre/ebooks/metadata/sources/identify.py
@@ -85,7 +85,11 @@ class ISBNMerge(object):
                 isbns, min_year = xisbn.get_isbn_pool(isbn)
                 if not isbns:
                     isbns = frozenset([isbn])
-                self.pools[isbns] = pool = (min_year, [])
+                if isbns in self.pools:
+                    # xISBN had a brain fart
+                    pool = self.pools[isbn]
+                else:
+                    self.pools[isbns] = pool = (min_year, [])
 
             if not self.pool_has_result_from_same_source(pool, result):
                 pool[1].append(result)
diff --git a/src/calibre/ebooks/metadata/xisbn.py b/src/calibre/ebooks/metadata/xisbn.py
index 56156c034e..ff21dfd89d 100644
--- a/src/calibre/ebooks/metadata/xisbn.py
+++ b/src/calibre/ebooks/metadata/xisbn.py
@@ -45,6 +45,11 @@ class xISBN(object):
                 ans.append(rec)
         return ans
 
+    def isbns_in_data(self, data):
+        for rec in data:
+            for i in rec.get('isbn', []):
+                yield i
+
     def get_data(self, isbn):
         isbn = self.purify(isbn)
         with self.lock:
@@ -57,9 +62,8 @@ class xISBN(object):
                     data = []
                 id_ = len(self._data)
                 self._data.append(data)
-                for rec in data:
-                    for i in rec.get('isbn', []):
-                        self._map[i] = id_
+                for i in self.isbns_in_data(data):
+                    self._map[i] = id_
                 self._map[isbn] = id_
             return self._data[self._map[isbn]]