Nik,
Are you trying to instrument a general case where the difference between control and variant is not limited to the client side stack? If server can be excluded from the experiment, then you’re in the traditional optimizational A/B testing space and tools like Visual Website Optimiser or Optimizely will be a good place to start. If, however, your change extends to the server stack, the only solution along the lines of what you’re talking about is Variant.
In general, you don’t want to expose the fact that there’s an experiment to the user by altering the URI (e.g. with a querystring parameter), but rather have these paths be private and server side forwarded to. Exposing variant URIs confuses search engines and alerts the user to the fact that he’s in experiment.