The Cloud Resilience Assumption Gap: A Taxonomy of Cloud-Native Failure Modes and Practitioner Validation
Abstract
Enterprise cloud migration is commonly justified by expectations of improved availability and operational resilience. Many organizations find that cloud transition does not automatically deliver those outcomes, even after adopting multi-zone deployments, managed services, and cloud-native recovery patterns. This paper identifies the structural root cause: inherited architectural assumptions formed in onpremises environments that conflict with how cloud systems actually fail. This paper defines this structural misalignment as the Cloud Resilience Assumption Gap. The paper introduces a taxonomy of five cloud-native enterprise failure modes: control plane failures, availability-zone coupling, managed service dependency cascades, configuration-induced failures, and edge infrastructure failures. Each category is derived from analysis of provider-published post-mortem documents and evaluated against twenty documented cross-provider incidents between 2019 and 2026. A longitudinal case study at a large-scale U.S. retail enterprise found that correcting the five inherited assumptions was associated with MTTR dropping from a mean of 5.4 hours to 43 minutes and a 60% reduction in change failure rate over a twelve-month period.