Agile Research - Tiger Teams and Other Explorations

A discussion around how we do research to support development and how this differs from normal development. Are research tasks something that get scheduled as normal tasks in a sprint? Do they need timeboxing more rigorously and what happens when the answer that comes out of the box is "I dunno". How do you constrain sloppy prototypes and code from playing with technologies from seeping into the main code base. Should you worry about this? Do you use different technologies, like scripting languages, for these kinds of explorations?