Taxonomic revisions are usually published as stable scholarly outputs even though the conclusions they contain arise from changing constellations of specimens, character observations, sequence data, analytical procedures, species hypotheses, and nomenclatural decisions. This mismatch makes it difficult to reconstruct why a taxonomic conclusion was reached, determine which evidence belonged to a particular revision state, or distinguish later reinterpretation from silent replacement. We argue that insect taxonomic revisions should instead be treated as versioned and reproducible research objects. The proposed architecture separates physical specimens and types, observations, character and genomic derivatives, analytical outputs, interpretations, taxonomic decisions, and nomenclatural acts while preserving explicit relationships among them. Evidence provenance records where an observation or derivative originated; decision provenance records how heterogeneous and sometimes discordant evidence was interpreted; version provenance records how matrices, genomic datasets, analyses, and hypotheses changed through time. This separation is important because coalescent groups, barcode divergences, gene trees, algorithmic clusters, morphology, behavior, life-stage characters, and genomic signals do not have interchangeable inferential meanings. The approach does not promise automatic species resolution or nomenclatural correctness. Rather, it is intended to make revision histories reconstructable, auditable, and capable of retaining rejected, superseded, and uncertain hypotheses alongside accepted ones. Implementation will require persistent identifiers, specimen-to-derivative linkage, explicit dataset versions, machine-readable relationships, and validation across contrasting taxonomic workflows.
