We investigate bubble dynamics in a submerged double-jet as an example of weakly coupled, Lagrangian transport. The setup resembles continuous casting of steel at Reynolds numbers Re=136000 to Re=272000. Using short time series of flow fields obtained from large-eddy simulations (LES) with a discrete bubble model, we calculate a database of transport patterns and time-extrapolate them to long durations in the framework of recurrence CFD (rCFD). A dedicated averaging procedure along bubble trajectories allows us to study their transport with large steps at little numerical costs and monitor their spatial distribution. Besides time-extrapolation for fixed Re with a single database, we demonstrate how several time series can be combined to (i) enforce symmetries and (ii) approximately model conditions in between. We compare bubble volume fraction fields and total hold-up obtained from LES and from rCFD and find very good to satisfying agreement with speed up factors of more than 500.