The rapid proliferation of mixed and augmented reality applications, latency-sensitive IoT systems, and time-critical cloud-native services has intensified the need for adaptive, resilient, and latency-aware microservice orchestration across heterogeneous cloud–edge environments. Emerging immersive applications such as augmented reality commerce and mixed reality surgical systems demand ultra-low latency, high availability, and robust fault tolerance, while simultaneously confronting network instability and resource heterogeneity.

This study proposes a comprehensive architectural and methodological framework for adaptive microservice orchestration and anomaly-resilient deployment across distributed cloud–edge infrastructures. Drawing exclusively upon established literature in microservices scheduling, edge deployment robustness, anomaly detection, Bayesian optimization, and mixed reality systems, the study synthesizes a unified theoretical model addressing latency-awareness, robustness, monitoring scalability, and intelligent anomaly detection.

A multi-layer adaptive orchestration framework is designed, integrating optimal cloudlet placement, robustness-oriented edge deployment, QoS-aware multi-resource management, latency-aware accelerator utilization, decentralized adaptive cloud–edge–dew models, hierarchical monitoring architectures, and machine learning-based anomaly detection. The methodology incorporates Bayesian optimization for orchestration parameter tuning, synthetic minority oversampling for imbalance mitigation, and hybrid anomaly detection pipelines using Random Forest, Support Vector Machines, and log-based failure prediction.

The proposed framework demonstrates improved theoretical latency resilience, enhanced deployment robustness, reduced service failure propagation, and improved anomaly detection reliability in heterogeneous environments. Conceptual evaluation based on industrial surveys, scheduling models, and architectural comparisons indicates improved system adaptability and fault containment compared to centralized orchestration approaches.

Conclusion: The integration of latency-aware orchestration, robustness-oriented deployment, and adaptive anomaly detection forms a foundational architecture for next-generation immersive and time-critical systems. The study contributes a theoretically grounded and practically aligned blueprint for scalable, resilient microservice ecosystems supporting mixed reality and distributed intelligent applications.