Take a slug of fluid, and assume: random particles (milli-slugs) across the slug will nucleate to vapor if Tfluid local is > boiling point at local pressure conditions. Note that the local expansion of these nucleate sites will raise the local pressure on surrounding areas of the droplet, inhibiting the boiling and expansion of nearby millislugs. Boiling extracts heat from the fluid, as does expansion of the vapor and this also changes local conditions of both fluid and surrounding gas environment. Modelling this way across several random or quasi-random distributions of the nucleation sites gives approximations of the slug/droplet breakup (shattering) that can then be correlated to high speed video or similar data.
A simpler approach not requiring the CFD and multiphase particle tracking implied in the above - ASSume only the outer layer of the slug/droplet vaporize first, and track the movement of the slug/droplet and evaporation over time/space. This gives an upper bound of the time or distance that the slug will stay liquid, since shattering (internal vapor bubble formation and expansion, with resulting breakup of the slug/droplet/stream/whatever) is not being explicitly modelled.
An even simpler approach - assuming a continuous stream of liquid enters the knockout tank, assume a chunk of it flashes entirely to steam, and then calculate its expansion within the tank and subsequent pressure rise of the tank. Add the next chunk of fluid, but now add a term for removal of some of the vapor in the tank due to venting and condensation...repeat ad infinitum until the flash event is done, or the tank explodes, or the system runs out of water...this could be a spreadsheet model. You could run it multiple times with smaller and smaller time step increments to see if it converges to an answer, and/or changing the assumptions of starting temperature and admittance/exhaust rates. Would fill an otherwise boring afternoon or two.
