<p>The extreme propagation environments of smart mining, characterized by heavy obstructions and dense deployments of collaborative devices, impose stringent requirements for ultra-reliable, low-latency, and fair connectivity. These conditions expose fundamental limitations in conventional wireless architectures, necessitating more flexible interference management. This paper proposes a unified resource allocation framework for smart mining cell-free networks assisted by a beyond-diagonal reconfigurable intelligent surface (BD-RIS) operating under an adaptive hybrid non-orthogonal multiple access (NOMA) and orthogonal multiple access (OMA) protocol. We formulate a joint max–min fairness (MMF) optimization problem that tightly couples discrete access point association, resource scheduling, access mode selection, and successive interference cancellation (SIC) decoding orders with continuous beamforming, power allocation, and unitary BD-RIS configuration across multiple physical timescales. To solve the resulting mixed-integer nonconvex program, we develop a safeguarded block coordinate descent (BCD) framework utilizing weighted minimum mean-square error (WMMSE) reformulations and Riemannian manifold optimization to ensure strict feasibility preservation and monotonic convergence. Numerical results demonstrate that the proposed design improves worst-user throughput by up to 194% compared to RIS-free systems and by 25%–72% over conventional diagonal RIS and OMA architectures. Furthermore, the framework adaptively scales NOMA utilization from 55% to 85% based on network load while maintaining SIC feasibility above 90%. These results confirm that BD-RIS-assisted hybrid access provides a robust fairness–efficiency tradeoff for mission-critical 6G smart mining networks.</p>