Public transport reliability is often undermined by two interrelated factors: in-vehicle crowding and headway variability. While both phenomena are widely recognized, limited empirical research has quantified their combined effect in high-frequency surface transport systems, particularly under conditions of overlapping routes. This study analyzes over 1.7 million stop-level observations from the Geneva bus and tram network, integrating Automatic Passenger Count (APC), Automatic Vehicle Location (AVL), and GTFS schedule data. A random-effects ordered logistic regression model is applied to estimate how operational irregularities and network design features shape crowding levels, measured through a five-level index. Results show that headway variability is the strongest predictor of in-vehicle crowding: vehicles arriving late absorb accumulated demand and are significantly more likely to exceed design capacity. Route overlap amplifies this effect by introducing interdependencies between lines, leading to higher crowding levels in shared segments. Vehicle type also plays a role, with trams more frequently operating under crowded conditions, while trolleybuses remain less affected. These findings provide actionable insights for operators: interventions such as holding control, coordinated scheduling across overlapping lines, and targeted stop-level monitoring can mitigate the most severe crowding events without major infrastructure investments. The study contributes to transport reliability research by providing large-scale empirical evidence of the crowding–headway–overlap nexus, and offers practical guidance for agencies aiming to improve comfort and service quality in urban surface transport systems.