The Forensic Residual Framework - Evidence-First Software Construction, Behavioral Reconstruction, and Deterministic Residual Endoduction from the DSFB Prior-Art Stack

Building trustworthy software is not principally a production problem. It is an evidence problem. This is most visible when reimplementing mature software, where a new implementation may compile, pass its own tests, conform to a written standard, and improve memory safety while still differing from the implementation that users, scripts, network peers, filesystems, package managers, operators, and downstream programs actually depend upon. The same problem also appears in new software whenever tests encode developer expectation rather than measured behavioral evidence. Differences often occur at surfaces that are easy to dismiss as incidental: diagnostic wording, exit status, ordering, malformed-input behavior, filesystem side effects, timing boundaries, platform behavior, wire representation, undocumented defaults, and version-specific historical behavior. This paper presents the Forensic Residual Framework (FRF), an evidence-first methodology for software construction, behavioral reconstruction, compatibility engineering, and verification-oriented development. FRF originates in the Drift-Slew Fusion Bootstrap (DSFB) prior-art stack, where residuals are treated as deterministic structural evidence rather than as noise to be averaged away. FRF treats observable disagreement between a declared authority and a candidate implementation as a residual that must be preserved and investigated before it may be normalized, dismissed, repaired, or accepted as an intentional divergence. Its central rule is residual primacy: a mismatch is not noise until evidence establishes why it is noise. FRF organizes verification work as authority admission, behavioral hypothesis, oracle experiment, raw capture, implementation, forensic parity court, residual creation, deterministic residual endoduction, classification, minimization, correction or explicit explanation, regression invariant, reproducible receipt, and bounded claim. The framework introduces narrow executable courts, typed residual taxonomies, authority-admission records, evidence-preserving normalization, minimized reproducers, explicit negative capabilities, authority separation, claim-precision ladders, parity ledgers, replayable receipts, and anti-drift gates. The contribution is methodological rather than a proof of universal equivalence. FRF provides a disciplined way to state exactly what has been observed, against which authority, under which conditions, what remains unknown, and what evidence would falsify the claim. Its objective is not to make software appear trustworthy. Its objective is to make trust inspectable. Keywords: software verification; DSFB; deterministic inference; densorial inference; tekmeric inference; software reimplementation; compatibility testing; differential testing; test oracles; behavioral parity; reproducible software engineering; residual analysis; legacy software; evidence engineering; deterministic endoduction.