RFC: Change rules for adding/changing test cases

[Joe Buck]

> > Unfortunately, there's no process in place to regulate the contents of
> > PRs.  

Richard Earnshaw writes:
> Then perhaps there should be one, well a process at least.  We have the 
> beginnings of one with rules for priorities etc.

We can't regulate the *initial* contents of the PR (since these come from
the general public), but we could put requirements on what must be added
to the report to mark it as fixed (examples: pointers to relevant mailing
list discussion and patches, explanation of what caused the failure, etc).
We should avoid making things too cumbersome, though.