Since you have to pass two (sometimes) conflicting tests, you get the worst of both systems. You get old-school managerial abuse and the careerist jockeying for visibility that occurs when large companies attempt to be "peer-driven". Instead of offering two paths to success, Google offers two orthogonal roadblocks, both of which must be overcome.
20% time is alive and well if you have a supportive (i.e. good) manager. If you have a typical manager, you can probably get away with 20%T as long as it's not obvious. If you have an evil manager, then watch the fuck out. There's a 5% cutoff for PIPs (the bureaucratic humiliation derived from the principle that it's better to keep a fired employee around for 2 months pissing all over morale than to cut a severance check) and having a 20%T project is an easy way to end up on the hitlist.
The danger of 20% time at a place like Google is that (a) the best work is awesome, but most of the work isn't very interesting, so (b) pretty much everyone is trying to get on a better project (and starts campaigning from about the 7th week) and (c) managers know this and don't like people who make it obvious. 20-percenters are either trying to transfer (that's the best use of 20%T) or to start something up within Google (which expresses the same flight risk, but is far less likely to work). So if a manager has to throw someone under the bus because a team is underperforming and doesn't have enough "calibration" points, 20-percenters (flight risks) are the first ones to go.