Consider the following: geometry is a bifurcation (Y) inlet: velocity prescribed. 2 outlets: where different resistances are applied R1 smaller than R2. The resistance b.c. is a simple P = Q * R condition where R is a representation for what happens downstream.

I can't get this converged. I get inflow on the outflow.

The resistance b.c. is just the simplest of lumped parameter models that I can think of, and it has a nice electrical analogy.

Can anyone shed some light as to why this is not converging, or give me an idea what to apply instead?

(1). Install two positive displacement pumps at each outlet. Then you can control the exact flow rate. And the outlet is always the outlet. This is a test setup. (2). With the flow rate specified, you can specify the outlet velocity as boundary conditions. The solution should converge. It is like having a faucet at the end of each outlet, you control the flow rate there.