<p>User scheduling methods in 5G and B5G networks are evolving from purely radio-centric optimizations to integral components of comprehensive network and system management strategies. The shift towards more dynamic, intelligent, and automated network operations necessitates user scheduling algorithms that are not only efficient in resource allocation but also adaptable to complex network architectures, capable of managing interference, and aligned with broader goals of automation and energy efficiency. In this paper, we have derived two closed-form expressions for the Signal-to-Interference-Plus-Noise Ratio (SINR) in extra-large scale MIMO (XL-MIMO) systems with spatially correlated multi-state line-of-sight (LoS)/non-LoS (NLoS) channels. One expression is asymptotic, valid for a sufficiently large number of antennas and in the absence of pilot reuse among scheduled users, while the other is ergodic, approximating the average SINR when the number of scheduled users exceeds the number of antennas serving each user. Numerical results validated the accuracy of both expressions, and we also provided simplified versions that demand significantly less computational effort. Furthermore, we proposed an efficient user scheduling (US) algorithm for crowded XL-MIMO systems that maximizes the sum spectral efficiency (SE) leveraging the derived analytical expressions instead of relying on instantaneous SE computations. This approach enables the US to remain valid over hundreds of channel coherence blocks, significantly reducing computational complexity. Our numerical results demonstrate that the proposed US algorithm substantially improves the sum-SE in crowded scenarios. By properly admitting pilot reuse, the algorithm effectively minimizes inter-user interference in XL-MIMO systems, mainly as different users are most likely served by a different set of antennas and thus interfere minimally with one another.</p>