Parametrized tests now really do ensure that the ids given to each input are unique - for example, a, a, a0 now results in a1, a2, a0 instead of the previous (buggy) a0, a1, a0. This necessarily means changing nodeids where these were previously colliding, and for readability adds an underscore when non-unique ids end in a number.