In the past years, the research and operation communities on climate, weather and hydrology have put efforts into developing on-demand services for the monitoring, forecasting or emergency response and recovery phases of extreme hydrometeorological events. This is the case for the ‘Copernicus EMS On Demand Mapping’ for natural disasters, including flood inundation, as well as the ‘Destination Earth’s on-demand extremes digital twin’ flagship initiative of the European Commission. These efforts often require new, configurable on-demand prediction capabilities to run Earth system models at very high resolution on global scales. From the hydrological sciences and services perspective, it raises questions about how the diversity of operational hydrological prediction systems that support local modelling and decision-making can integrate this new paradigm, without losing efficiency and predictive accuracy in the process. In this study, we investigate existing (or soon-to-be) operational flood impact modelling simulation capabilities in nine countries: Bulgaria, Czech Republic, Denmark, Finland, France, Iceland, Ireland, Slovakia and Sweden. We developed technical model workflows for each country to illustrate the diversity of approaches encountered in national flood forecasting systems. Each workflow is a visual diagram that identifies nodes represented by start/end points, and tasks and processes that affect the outcomes (i.e., the flood forecasts). Workflow developers were guided to reflect on aspects such as offline setups (domain discretization, model calibration), input data (acquisition, type, source), data pre-processing steps, models and associated routines (data assimilation, post-processing), and outputs (web-based interfaces, visualization). Guidance for inter-comparable workflows were discussed, which allowed us to reflect on a generic workflow to depict the way data and models interact in the context of flood forecasting and warning. Altogether, the hydrological/flood forecasting technical workflows highlight the needs of each configured system to locally pre-process meteorological data before using them as input to the hydrological models. This may include different actions: file reading, data formatting, data interpolation, computation of sub-catchment areal precipitation, etc. As the workflows rely on continuous hydrological modelling (as opposed to event-based models), the role of model initialization to capture the catchment initial conditions at the time a forecast is issued (e.g., the amount of water stored or flowing in the catchment before a flood event) is also highlighted. These are important aspects to be considered when interfacing national flood forecasting systems with continental or global on-demand services. The workflows offer a comprehensive and diverse view of the many components that can facilitate or hinder reproducibility, transferability, and (event- or user-driven) triggering of on-demand services, contributing to inform the setup of new approaches that aim at more interactive and configurable access to data for flood risk assessment at different scales. This work is funded by the EU under agreement DE_330_MF between ECMWF and Météo-France. The on-demand capability proposed by the Météo-France led international partnership is a key component of the weather-induced extremes digital twin, which ECMWF will deliver in the first phase of Destination Earth, launched by the EC.