In this paper, Aaronson and Arkhipov use the following lemma (lemma 67 in the paper):
Let $V \in \mathbb{C}^{n \times n}$ be a matrix of rank $k$. Then $\operatorname{Per}(V + I)$ is computable exactly in $n^{O(k)}$ time.
Unfortunately, they refer to a "forthcoming paper" which I was unable to find. Is there a publication where this or a similar statement is proven?