The physical tests have some known issues and room for improvement.

• the point state is not recorded correctly: the point is switched and immediately afterwards its state is sampled. At that point, we haven't had time to receive the feedback from the corresponding control board on the execution state of the point.
• memory is not always freed correctly.
• tests with train driving: after we set the points, we don't check if they are in the expected position -> if a point is not working correctly, we'll likely get a short circuit.
• we could use some more documentation and perhaps debugging info, for example the track output and booster state when the SIGINT callback is triggered, i.e., if we manually abort a test.

Work on this has already begun in branch fix-physical-test.