> It would be nice if we could store tests.
>   .be/BUGDIR/tests/...
> and link them from bugs.

Better: have them be comments with a TEST tag.

The mime type could hint at the execution mechanism:
  text/x-python
  application/x-sh
  ...

> Then running
>   test.py BUGDIR/BUG
> would run the tests for that particular bug.
> 
> This would provide regression testing via
>   test.py $(be list --ids --status fixed)

This should be a 'test' command (libbe.command.test.Test), since
people will want to test bugs for their own projects, and out current
test.py is for testing BE specifically.  It should be
  be test BUGDIR/BUG
  be test $(be list --ids --status fixed)

We _should_ add be
  test $(be list --ids --status fixed)
to test.py for regression testing.

This whole thing would make the fixed/closed distinction more clear,
since fixed bugs would get tests run and expect success, while closed
bugs' tests would be skipped.

Finally, if users are submitting tests on their own, it would be a
good idea to sandbox them, but a portable way for sandboxing scripts
sounds very complicated.  It would probably be easier to sandbox
python scripts, but I don't know what that would look like...

A work around would be to allow users to post tests, but not allow
them to set the TEST flag.  Then the bugdir maintainer could set the
flag themselves once they'd vetted the test.  Much uglier than
sandboxing, but also much more easily implemented.