Answer
Check the keep-alive and request-timeout parameters configured on the middleware server before assuming a network or client-side cause — increasing these timeout values on the MWS has resolved this class of intermittent, seemingly-random connection-timeout error for some environments.