This paper introduces a novel variational approximation for inference in (inverse) Wishart processes. The reviewers debated this work extensively in the post-rebuttal stage. They were worried about an empirical focus on specific financial modelling applications, but agreed that the paper offers interesting insights into the more general problem of inferring covariance matrices. Overall, I agree with the reviewers that it offers insights of interest to a significant sub-community at NeurIPS, and thus can be accepted to the conference