Output-Specific Numerical Auditing of an Idealized Bonded-Block Model: Compression and One-Way Crusher-Load Replay

Agreement in a global reaction does not necessarily preserve local fracture outputs. We audit this distinction in idealized bonded blocks using a four-mode shared-scalar damage law, with separate independent-mode comparisons and prescribed crusher-load replay. Parameters are numerical construction values. Across five screened 125-sphere topologies, three time steps retained first-failure identity, trigger mode, and immediate connectivity. A separate damaged baseline showed close sampled mean reactions but different failed-bond sets in a four-step post-failure comparison. Additional velocity perturbations as small as one trillionth of the inherited root-mean-square (RMS) velocity also changed a near-threshold bond outcome at a fixed time step, whereas neighboring damping coefficients changed component membership. These controls prevent attributing the baseline differences uniquely to temporal discretization. In a two-block one-way replay, the prescribed input histories were identical; one block retained its terminal fracture structure and the other did not. Small work-ledger residuals, sampled peak agreement, event identity, and component membership therefore answer different numerical questions. The results define a model-specific audit of these outputs under stated loading and damping choices.