Unplanned train disruptions are a source of passenger dissatisfaction because they are often accompanied by overcrowding and lack of information. To better accommodate passengers during disruptions and preventing travellers from switching to other less sustainable modes of transport, mitigating control strategies can be applied by railway operators. This however requires predicted passenger flows over all available travel options as an input. Due to the COVID-19 pandemic these passenger flows have becomes less predictable, as many travellers have gained an additional feasible alternative to cope with unplanned disruptions on outbound commuter trips − they may return home and start teleworking. Because this travel option is only available to teleworkers and now utilized more than before the COVID-19 pandemic, heterogeneity in route choice behaviour has increased. To fill this knowledge gap and provide predictions of passenger flows, an online survey containing a labelled stated choice experiment was carried out among Dutch train commuters. Consequently, a latent class choice model was estimated to investigate the influence of disruption characteristics, teleworking, COVID-19 risk perception and information provision on travel behaviour during train disruptions in the Netherlands and uncover heterogeneity in behaviour. Our results indicate that the strongest predictors of route choice behaviour are the moment of discovering the disruption, the disruption length and job characteristics. Uncovering four latent classes shows the different valuations of crowding, waiting times and additional travel times among commuters. Commuters with the option to telework are more likely to return back home during disruptions as well as commuters who are sceptic towards the provided information and those who are still conscious of COVID-19. Commuters who cannot telework and trust the provided information are more likely to reroute within the train network whereas commuters who cannot telework and do not trust the provided information are more likely to wait for the disrupted services to resume.