I've discussed this with a few people and have been reminded of a Joel Spolsky interview - the grandmother story ("the process is overtuned") on page 2 caught my eye. Basically, he explains that over the years Microsoft has picked up so many good habits (by correcting their mistakes) that the process (testing, validation and so on) has become more important than the actual *thing* - the product being created.
When you're on a 6 month cycle, you're in feature freeze for 3 months. That's 6 months a year that we're not focused on new features.
Now, if you move to a 9 month cycle, you still have a 3 month feature freeze per cycle, but you're only spending 4 months a year in freeze, so you've gained 2 months innovation/breaking time. Experiments can happen right in GNOME.
If we stay in the 6 month cycle, the breakages happens outside the main GNOME tree (as Luis said). That's risky for some young gun to take on in his garage.
As an alternative to the "clean break", the leap of faith that Luis is talking about, perhaps we can just have a little more madness in GNOME devel. Doubling the feature addition period, moving the focus from testing and bug fixing to innovation for 1 quarter a year more, might be a way to do that.