End-edge-cloud orchestration has become a fundamental paradigm for distributed video analytics (VA), enabling intelligent coordination of computation and communication resources across heterogeneous layers. The orchestration framework has to address several intertwined challenges, including time-varying video content, fluctuating network conditions, and the trade-offs among inference accuracy, latency, and energy efficiency. Adaptive configuration mechanisms dynamically tune frame resolution, sampling rate, and neural network (NN) model selection according to contextual variations in content and bandwidth. Temporal and spatial video pruning techniques enhance transmission efficiency by selectively retaining frames and regions of interest, mitigating redundancy while preserving analytic fidelity. Neural network partitioning further distributes inference workloads among end devices, edge servers, and cloud centers, optimizing the placement of layers to balance computational and communication costs through profiling, graph-based, and online decision frameworks. Multi-edge query scheduling extends this collaboration to large-scale deployments, coordinating workloads across multiple edge nodes via centralized optimization or distributed multi-agent reinforcement learning, thereby improving scalability and responsiveness.