Temporal Layering and Rule-Based Matching in Distributed Settlement Ledgers: A Multi-Horizon Reconciliation Architecture for Financial Break Detection

Financial settlement breaks remain a persistent operational risk. Batch-only reconciliation detects them only after end-of-day cutoff, while stream-based approaches often sacrifice coverage or lack shared rule semantics across horizons. This paper presents a multi-horizon reconciliation architecture coordinating three tiers over one transaction stream: daily batch, hourly incremental partitioning, and real-time listener-based interception. The tiers share a governed SQL rule repository and an exception store that prevents duplicate alerting. We formalize reconciliation as a matching problem, define keys, discrepancy categories, tolerance-based conditions, and exception classifications, and give a matching algorithm with representative SQL rules. A case study on a distributed loan origination system examines four discrepancy categories: compatibility processing errors (210), suspension events (45), classification anomalies (1,200), and calculation discrepancies (85). The contributions are coordinated three-tier design, shared rule governance, and evaluation of coverage, latency, and maintenance trade-offs. Measured results are distinguished from design targets; no unmeasured performance improvement is claimed.