Robust and on-time airport operations are key to safe and efficient air transportation. Hence, statistics show that airport delay is one of the dominant contributors to reduced system performance, with service disruptions resulting from flight delays and cancellations imposing significant economic impacts to the stakeholders and passengers. Predicting airport related disruptions is however difficult to model in the absence of comprehensive and reliable data sources. Based on an analytical disruption model, this work presents a scalable, machine-learned methodology for quantifying and classifying airport disruption dynamics, enabling the identification of recurring patterns of performance degradation and recovery without relying on detailed causal data. The methodology is applied to a case study comprising data of a large set of US airports, extracting context-specific disruption dynamics through cluster analysis.