Isolate backend state when running multiple agents

[jnicklas]

For applications where we do not mock the backend, when using multiple agents, our current strategy of running all agents in parallel is problematic. Since the backend will maintain state, multiple tests running at the same time becomes problematic, because the backend state will not be isolated from each other.

As far as I can see, there are two approaches to solving this problem:

• Start an app server for each agent. This seems a bit crazy to me.
• Run agents sequentially, rather than in parallel