Considerable efforts have been made towards the development of robust and fully automated vibration-based monitoring systems with the goal of extracting relevant information regarding the dynamic behaviour and health condition of the monitored structure. Satisfactory results have been obtained in this scientific domain by combining accurate measurement systems and fully automated output-only modal identification techniques. In this context, the main objective of this contribution is to demonstrate in a full-scale case study that with the application of sophisticated algorithms for the automatic tracking of modal parameters, it is possible to detect very small structural changes. Apart from describing the main features and capabilities of the autonomous monitoring system implemented to assess the structural condition of a peculiar football stadium suspension roof, this paper also outlines the main results obtained over the course of 4years of monitoring carried out to assess the dynamic behaviour and the health condition of the roof structure. The routines developed for the online processing of the continuously collected acceleration time series include state-of-the-art processing techniques, such as automated modal identification based on cluster analysis and principal components analyses combined with control charts for removal of environmental or operational effects and detection of structural changes, together with some innovative features in the context of continuous dynamic monitoring, such as the quantification of the uncertainties associated with each modal estimate and the estimation of the contribution of each mode to the measured structural response.