What happens if you have to get something right? It might be safety critical (people die if you get it wrong) or it might be business critical (if you get it wrong you lose money) or it might simply be that there are multiple different ways of deriving the answer and this helps you make smarter decisions (if you get it more right you make more money).
This is where the Cheshire Cat comes into play in that "what I tell you three times is true". These are cases where having redundancy is not about system reliability its actually about the reliability and performance of the developers of the software. Most of the time its not worth the effort or cost to do this (Boeing famously decided for the 777 that three independent primary flight control systems was too expensive) but in an SOA enabled business it is worth thinking about areas where it could be beneficial to have software duplication.
Areas that would suit this approach in a normal business need to have the following characteristics
- High value or High impact - if you get it wrong (or less accurate) there is a business case for improvement
- Its not about reliability - that is a hardware and infrastructure problem
- There are multiple ways to skin the cat - It isn't a case (ala flight controllers) of one piece of software in three different teams, its actually three different approaches to solving the problem
- There is a way to judge between the answers
Forecasting is one example that could suit a duplicate development approach with the "vote" going based on historical accuracy and consistency of trends.
The simple point is that as SOA takes off in businesses its time to look at investing as well as saving money, and sometimes investment might mean doing something more than once for the right reasons.