Paleomagnetic data processing involves a hierarchical chain. Classically, uncertainties are being calculated at a single level only, which is then taken to represent the entire chain. Justification for doing so is that the secular variation of the field is larger than the measurement uncertainty. Clearly more formal error propagation would lead to more robustly constrained interpretations, which would be so important for paleogeographic reconstructions, for example. However, this is more insidious than one would like as a required transformation inherently distorts the data distribution. Heslop and Roberts  have now formulated a rather straightforward solution to this problem that importantly can also be applied in retrospect to legacy data.
Citation: Heslop, D., & Roberts, A. P. . Uncertainty propagation in hierarchical paleomagnetic reconstructions. Journal of Geophysical Research: Solid Earth, 125, e2020JB019488. https://doi.org/10.1029/2020JB019488
—Mark J. Dekkers, Associate Editor, JGR: Solid Earth