Imagine for a moment that there were no tests that failed. Does that mean that you'd have no category for those tests or that the category would consist of a null set: "{0}"?
If you're talking about the column titles for a table, then perhaps the titles would be better as:
| *TESTS* |
------------------------------
| *Failed* | *Passed* |
------------------------------
0 #1
#2
#3
The category name tells the reader what kind of members the category does or should contain, regardless of whether the category is empty, has only one member, or has multiple members.
The "Test which failed" category is just like headline English: it isn't subject to normal grammatical rules, but you seem to be judging it that way. In addition, it's either a question-begging category (you knew before you did the tests that you wanted one test to fail, so you labeled one category to indicate the outcome you wanted and then you ensured that this was the case) or an ill-worded category: the two categories can more briefly be named "Passed Tests" and "Failed Tests", which eliminates the need and, one hopes, the misguided desire for number agreement between the category name and the number of items it contains.
If you're writing sentences, then you've got to say something like "One test failed, but two tests passed".