Website monitoring company Net Applications has discovered that around 30% of the traffic originating from Google HQ has identifying information stripped from the user agent string.
This traffic, which people are assuming is Google employees, has the operating system identifier stripped from the user agent string.
Speculation is that Google is testing a Google OS amongst employees:
Whilst this might be true there are other explanations for the data. My first thought was that Google was using a stealth spider pretending to be a normal user to build up a more accurate picture of the web. Googlebot can’t see images, scripts, Flash and accept cookies. A stealth Googlebot could do all of these things and help build a better index in the future.