It seems as if allowing this would speed up testcase correction and reduce comments.
Well, it depends.
If you are given a certain protocol to be applied to a certain input, then it is fine to edit if the community spots the error.
So, if your challenge is to output
n, you can fix it.
If the output is interpretable (like an ASCII-art challenge), you better should ask the author to either make clarifications or correct the test case. You may edit that, and if you are proved to have misedited, anyone can call a rollback to the initial test cases.