This reminds me of the whole agent thing -- pretty much dominated by Java (e.g., Jade, Jason, Jack) nowadays --, for which I would bet lots things are done more straigthforward using Haskell -- especially those parts the Java coders are usually proud of... Let's maybe speak of *second order scalability*:

As first order scalability would rather be a matter in space time load increased by repetitions, the concern of second order scalability would be more about a *fractal* expansion of concepts like a *closure* -- Haskell, already in a vivid exchange with interactive theorem proving (e.g. Coq adopts type classes from Haskell and dependent types vice versa) seems excellently prepared... :-)

Interesting. I´m working in something like second order scalability. Instead of brute performance by  redundancy,  high speed networks and fast disks, scalability can be achieved by looking at the properties of the data. 

I ever tended to say financial applications are especially prone to be boring -- the prototype of repetitive IT, even for strategy the stupid 'traffic lights cockpits' or OLAP(!) ... But this problem is rather supply driven to me.

For sure. This is supply driven. There are a lack of new ideas mainly because the technology is low level and obsolete.