Error detection and fault localization in microservices with per-service metrics and baseline-calibrated thresholds

Reproducibility package accompanying the article of the same title. The proposed Three-Signal Hybrid combines per-service metrics, distributed traces and Kubernetes symptom events to detect errors and rank suspected faulty services. Detection thresholds are calibrated using normal system operation. The package contains: Method implementation and dataset adapters. Experimental protocols, amendments and version history. Evaluation scripts, comparators and robustness analyses. Dependency specifications and reproduction instructions. Derived evidence and results for RCAEval, OpenRCA Market and TeaStore. Raw TeaStore telemetry, including metrics, traces and Kubernetes events. Raw RCAEval and OpenRCA datasets are not redistributed; their sources and input checksums are documented. The package also describes methodological limitations, including the retrospective reconstruction of aggregated Kubernetes events and the distinction between window-based replay and continuous online operation.