We do distribute and sell independently of the Mac App Store, which allows us to offer trials, lets users to easily convert the trial to a licensed copy, and provide bulk or educational discounts. Even with that, ~80% of our revenue has come from the Mac App Store. Users want and expect software to be available on the Mac App Store; I'm sure the support volume we'd get by not being there would far outweigh any we've seen from this bug.
Sandboxing was a technology that did not exist (and therefore was not required) when we started selling on the Mac App Store. This came later, and became a requirement to push any updates containing features.
Lots of mac developers have complained publicly (or privately to me) about sandboxing. Many others have not been able to issue updates, removed major features, or simply removed their apps from the store. There's plenty of information on this on blogs or the developer forums, though many would not speak out publicly against the hand that feeds them. The apps that do well with sandboxing are naturally those which do the least. It would have been easier for us to neuter our app, but we've taken a strong approach to working through all the issues and avoided removing any major features. It would be unfortunate if innovation was sacrificed for security.
We originally planned to have our non-MAS copy not be sandboxed. This creates complications and other bugs for users when offering both versions based on how the sandbox container is maintained and that any bookmarks (aliases) kept cannot interoperate between the two. From all the tradeoffs we see, fully adopting sandboxing is the best option. But, that won't stop me from complaining about all the bugs!