Closedness of bounded strength and partition rank loci

We show the existence of a quartic form of strength $3$ in the closure of the strength at most $2$ locus, answering a question of Ballico–Bik–Oneto–Ventura in the negative. In the analogous setting of partition rank as introduced by Naslund, we further show that the locus of tensors of partition rank at most $2$ is closed, but the locus of tensors of partition rank at most $3$ is not. Together with a suspension lemma, this answers for all $r$ whether the locus of forms/tensors of strength/partition rank at most $r$ is Zariski-closed.