The paper tackles a fundamental question about certain ways of measuring the "complexity" of mathematical objects called multilinear forms and tensors. These are generalizations of matrices to higher dimensions, and researchers want to understand how to classify them by their structural complexity. Two specific measures of complexity are studied here: "strength" (for polynomial forms) and "partition rank" (for tensors). The central question is whether the collection of all forms or tensors with complexity at most some fixed number r forms a "closed" set, meaning roughly that if you can approximate an object arbitrarily well by things of complexity at most r, does the limit itself also have complexity at most r?
The main finding is that the answer is no for strength: the paper constructs an explicit example of a degree-four polynomial that has strength exactly 3, yet it can be approximated arbitrarily closely by polynomials of strength at most 2. This settles a previously open question by showing that being "low strength" is not a stable property under limits. For partition rank, the situation turns out to be more nuanced: tensors of partition rank at most 2 do form a closed set, but tensors of partition rank at most 3 do not. Again, an explicit counterexample is constructed to demonstrate this failure of closedness.
By combining these specific results with a general technical tool they call a "suspension lemma," the authors are able to extend their conclusions to all values of r simultaneously, not just r equal to 2 or 3. This gives a complete and clean answer to the original question for every possible complexity threshold: for strength, closedness fails for r of at least 2, and for partition rank, closedness holds for r of at most 2 but fails for r of at least 3. The work settles a natural and previously unresolved problem at the intersection of algebraic geometry, combinatorics, and the study of tensors.