Regardless of whether we fix the memory leak or buy more RAM we are still running our production service on a single server, and could do with a high-availability setup.