We saw a weird crash of a test run once in _update_cache() where

self._tbuf was unexpectedly None.  The only way this can happen is
when the storage is closed (probably by a different thread); close()
sets _tbuf to None.  It turns out that a TransactionBuffer instance
can safely be closed more than once, so there's no need to set _tbuf
to None in the close() method.

src/ZEO/ClientStorage.py

@@ -266,9 +266,7 @@ class ClientStorage:
 
     def close(self):
         """Storage API: finalize the storage, releasing external resources."""
-        if self._tbuf is not None:
-            self._tbuf.close()
-            self._tbuf = None
+        self._tbuf.close()
         if self._cache is not None:
             self._cache.close()
             self._cache = None