OK. I suspect it's a solver glitch. This assumes you didn't do anything with the Equations under Solver. Because there are no boundaries there are no momentum sources, so a zero flow is not unreasonable. Because there is then no disturbance in the flow there is no "error" to trigger the solver to check and iterate to improve the "error". So the flow remains at zero velocity. This continues. Instead initialise with a tiny cross flow (1e-5m/s is enough) and see how it behaves.
Note, because all four walls are the same the flow may be very chaotic.