Sometimes a given test function is invoked with multiple data tags that are not expected to reflect significant differences in performance in practice (i.e. a performance regression on one data tag is also reflected in all the others).

One reason for having many data tags for a given test function could be to demonstrate the time complexity of an algorithm (http://en.wikipedia.org/wiki/Big_Oh). In that case it makes little sense spending resources tracking all of the input sizes individually.

This should be cleaned up so that only one data tag (or a small representative set) is tracked for performance.