Merge the current state of the art of Felix's and my work on test structure improvements
This includes full randomization support and incomplete support (GUI stuff
missing) for keeping test run data consistent even if the structure of a test
changes afterwards.