Here's an idea - link it to your performance monitoring software of choice, and have it colour-coded to show the usage of your cluster nodes in real time. How cool would that be?!
Test frameworks use some pretty crazy tricks to produce a pleasant syntax with good error debugging messages. For an example, see CATCH: https://github.com/philsquared/Catch/wiki
I'd imagine this sort of stuff is much harder to port across systems than regular production code.