2. Your starter project will probably be pretty boring. Suck it up and do a bang-up job on it. Starter projects are usually a way for your team to get a sense of what you can do. If you establish a reputation as someone who gets shit done early on, good projects will come your way.
3. Stay away from - well, I can't exactly say in a public forum, but basically any mailing list or internal app where you get long centithreads or pictures of cats. Corp G+, however, is pretty useful for getting plugged in to what the general zeitgeist at the company is - what concerns the rank & file employees have, what projects people are excited about, interesting new infrastructure that's available, teams you may want to join.
4. 20% time is a great tool for getting to know a potential team and letting them get to know you, and a terrible tool for starting something new. When you first start, you should probably be heads-down on your main project, and dive in there. But once you've got a chance to see some of the other interesting projects in the company, don't be afraid to 20% there and transfer if you like the experience.
5. Find and attach yourself to a good manager, do good work for him, and ride the coattails of shared success. Managers vary wildly in quality at Google. Experienced engineers (those with >1 year of experience in a department) all know who the good ones are. If you get a bad manager you will stagnate, find your projects are continually ripped out from under you, and come up short at promo time. If you get a good manager your career will be one success after another, and eventually you'll get a reputation as the "go to" guy in a domain. That's when you get your pick of high-profile projects.