Hi Alex,

Can you please take a look at this one and let me know if you see any problems?

When I'm testing manually, I tend to run "run.py -v 2", and it has long annoyed me that failing tests are reported twice: once for the real failure, and once because they write an exception trace to stderr. This is my attempt to fix that, but I haven't checked it with other verbosity levels.

Thanks!
Michael.