We have got at least 3 problems that I recall in this relatively short period: wallpapers with Nintendo copyrights, Radio FM messing with speakers affecting phone calls and Load applet advertising screencasts that actually won't work, confusing several users.

Actually these three problems were not difficult to find, and they were even commented during the QA process. Perhaps what happens is that the goal is difficult to achieve in practice (one person going through 10 blocker criteria, with the skills it implies). Which leads to even the more riguroous evaluators being soft here and there.

I think the key feature that would make testing more accessible to power users and even fun is to split ratings by blocker criteria, as discussed some time ago. Some users will be happy checking whether the features advertized are in place. Others will follow the steps to check whether the app is optified. If at the end the remaining items are system performance and power managent, normal users will be able to assess at least if they noticed serious/noticeable problems. Of course much better if someone can run the right tools and do proper tests, but in their absence happy testers have chances to lead to happy users.