Recovering Process Boundaries from Complete Event Sequences in Finite Workflow Models

A complete ordered event record can leave the boundaries between named processes unresolved. This paper examines that problem in finite workflow models where each process type has a nonempty event pattern. It distinguishes grouped histories, flattened event words, composite effects, and evidence of actual occurrence. Exact process-type recovery is characterized by injectivity of the flattening map on a declared admissible language. For unrestricted concatenations, the paper reconstructs the classical Sardinas–Patterson residual criterion with explicit witness invariants, soundness, completeness, and termination arguments. A fictional document-handling dossier shows that two different process decompositions can preserve the same total process count, making recovery query-specific. A further example separates unique decoding of completed finite records from boundary identification during an ongoing stream. The discussion examines delimiters, identifiers, constrained histories, uncertain coverage, and competing criteria of process unity. Verified antecedents in coding, process mining, provenance representation, and experiential philosophy locate the proposal without a claim of conceptual priority. The formal results remain conditional on the vocabulary and record convention; historical truth, institutional boundary legitimacy, and responsibility require additional grounds. Open questions concern constrained languages, closure-sensitive inference, and comparison of process-individuation criteria.