Thanks for your reply Rob.
I'm using the Mixture model (explicit). I tried Eulerian first, but I found much more severe divergence issues with Eulerian.
I'm using the species-mass-transfer model for mass transfer. The from-phase and to-phase mass transfer coefficients are both calculated using the ranz-marshall correlation.
Some more info about my model, if needed: realizable k-e, timesteps ~1e-2 - 1e-3, domain is 1m x 1m with uniform 5mm mesh resolution. I've attempted lower timesteps and finer mesh, but it didn't seem to help.