The process of coalescence mist filtration in liquid- or gas-liquid systems is strongly controlled by the dynamics of the multi-component fluid transport at the pore- or fibrescales and its interactions with the filter media. However, current designs of mist filters are largely based on empirical data or on single-fibre filtration theory, primarily because of the complexity and difficulty in making accurate measurements at such (small) length scales. Current advancements in high performance computing provide a unique possibility to understand the dynamics of such flows using highly resolved droplet and interface capturing computational fluid dynamics (CFD) simulations - which can provide vital data for application-specific optimization of filter media. However it is important that the spatio-temporal resolutions required to accurately numerical model the fluid dynamics of micro or nano-fibre filtration processes at full size of the filters may typically demand simulations to be run with several hundred million (to over a billion) computational cells and long run-times. Hence, for reduced design lead times as well as computational co ... mehrst, it is desirable to keep the size of the filter domain to a minimum, while ensuring that the largest fluid structures and scales are captured in the simulations. A review of the (limited) literature on CFD simulations of the mist-filtration process reveals that the size of the filtration domain have been predominantly chosen rather arbitrarily based on the a set multiple of the Brinkman screening length or by the computing power available - and no reported studies are yet available that address conditions of high levels of fluid saturation that involve large fluid structures. In the present study, a series of systematic computational simulations using successively larger domain sizes are carried out to identify the relationship between the characteristics of the two phase flow (such as saturation, thickness of liquid layer, pressure drop, etc.) and the size of the filter domains considered. Two vastly different types of filter media, nonwoven and foam, with similar properties such as packing density, fibre (or element of foam) diameters are chosen to additionally infer the influence of filter structures at the pore-scales to the domain size. The transient simulations are carried out using the interface capturing volume-of-fluid (VOF) solver available within the open-source CFD framework OpenFOAM.