That sounds almost like a trick question, if I understand it correctly. If each row may be used only once, then the rows should be specific for a test and not in a common sheet (do you see the contradiction? ;).
Having a common set of data for multiple tests can be fine if all tests are allowed to that same data (e.g. default values for an address). This simply sounds like bad design.
Now, it is possible (if it wasn't a trick question) that the intent is to add some data variability to the tests (e.g. 10 sets of data, randomly divided among 10 tests every time). If that's the case, the data is either consistent (if it influences the test outcome, e.g. discount values for an order) or not important (if it doesn't influence the test outcome, e.g. a street name).
In both of these cases, you should use existing 'random data generators' that return random data and/or write your own where necessary.