Ian, without dishing out any personal insults or anything, what it boils down to is that every (supplier|vendor) can make mistakes and have problems and therefore it is every (user|client|customer)’s duty to think about how that will affect their business.
So yes I would say it is incredibly unwise to store critical data in any one place that is so far removed from your own control. The beta webmail service of a large search engine company would be just one example of a bad place to keep the one copy of your vital data.
What is so very different about this example and say, buying all your Internet transit from one NSP? If people being able to reach your network is essential to your business and you buy all your connectivity from one NSP whose network then dies, you can sit on a blog bitching until you are blue in the face about how the NSP should have forseen this and built a redundant enough network to survive it, etc. etc., but in the real world any network engineer will tell you that you should have had multiple upstreams for redundancy purposes.
Equally someone who put a lot of business critical data in one remote place hosted by some other company (especially given it is likely that their data and their custom means very little to that company) made a foolish choice.
Unless redundancy and local control is inherently built into the service then it will never be suitable for (power users|big business), and this is one major reason why I will never use gmail.