It seem to be logged for the incoming requests, most likely to later identify and fix the issues with the service. So, by sending random-random user agents you would make their work a little harder. You might also get banned for not complying with the service requirements. I suppose a random string is ok for one off testing, but for a regularly used script or a tool you better construct an actual user-agent string that identifies it.