I'm just trying to emulate mentally the situation:

2009-09-01: Developer promoted App v1.0 to extras-testing
2009-09-05: App v1.0 has got only 5 positive votes. Mixed feedback. Waiting.
2009-09-06: Developer promotes v1.1 with changelog, based on feedback received.

GA scenario (correct me if I misunderstood it):

One admin needs to look at the update and decide whether the 5 votes are still good or new review is needed? looks like a confusing situation with many possible variants

qgil scenario:

Counter resets. v1.0 is forgotten and v1.1 is now under review. The 5 voters look the changelog and check the app only to confirm their previous +1. Others can now check that the issues are addressed and vote. If v1.1 is good it will go to Extras in 10 days.

If the developer doesn't want to stop v1.0 because it looks like it will gather enough support within the first 10 days then he can just keep polishing v1.1 and release it once v1.0 has made it to extras. We are talking about only 10 days between stable releases! That is already fast compared to anything you regularly see even in the Linux desktop.

And just like it happens in the Linux Desktop, if power users want to get the fresh meat sooner and more often they can always decide to move to -testing or -devel. But it's good that average end users are kept away from so frequent updates. Specially considering that such fast iterations carry always a risk of regressions.