I just designed and built (and manage) a cloud-based virtual platform and while some issues are valid others are really a result of poor management.

For example, to delete an entire account is just poor management and poor validation of process.

But updating 50+ web servers can be a challenge. The example provided is great if all of your web/app servers are identical. Well how about when they are all completely different from one another but still house the same application (each one is for a different client)? Well our solution was to write a new management console specifically for our app a la System Center - so we can look out and see all instances of the app and update them one at a time or in groups or all at once just like System Center can. It is a unique solution but it works.

And what if you have flaws in the vendor's platform or software (ie. VMware vCloud has some flaws I am waiting for them to fix)? Using the Cloud requires thinking outside of the box and finding new ways to do things as well as compensate for items beyond your control. It's do-able and can be worth the effort but it requires planning and thought put into every single step - no rushing into anything.