Python 3: http.client.HTTPResponse reads too much

The Python 3.6 documentation for HTTPConnection.get_response() has this
note:

    Note that you must have read the whole response before you can send
    a new request to the server.

As far as I understand, the problem is that HTTPResponse() wraps the raw
socket object in a buffered reader that reads ahead, so the first
response object gets all of the pipelined responses into the buffer of
the socket wrapper that is discarded and not passed over to subsequent
responses.

src/zope/server/http/tests/test_httpserver.py

@@ -162,20 +162,20 @@ def testPipelining(self):
              b"Content-Length: %d\r\n"
              b"\r\n"
              b"%s")
-        to_send = b''
+        to_send = []
         count = 25
         for n in range(count):
             body = b"Response #%d\r\n" % (n + 1)
             if n + 1 < count:
                 conn = b'keep-alive'
             else:
                 conn = b'close'
-            to_send += s % (conn, len(body), body)
+            to_send.append(s % (conn, len(body), body))
 
         sock = socket.socket(socket.AF_INET, socket.SOCK_STREAM)
         sock.connect((LOCALHOST, self.port))
-        sock.send(to_send)
-        for n in range(count):
+        for n, req in enumerate(to_send):
+            sock.send(req)
             expect_body = b"Response #%d\r\n" % (n + 1)
             response = ClientHTTPResponse(sock)
             response.begin()