Doesn't really matter if you see the survey or not - valve can validate their data other ways. They easily know how many clients connect from each OS and what proportions as that's fundamental to the client itself. The survey fills in the rest of the data like which kernel, distros, and hardware.
All this would do is maybe weight some of the answers on which flavour of Linux and which hardware is being used in the favour of proactive users. But really good survey data relies on being representative and that is bes achieved by large random samples rather than people saying "count me!"