You can now sit down at a terminal with a fabric problem and know exactly what to type.
You know the device map: which commands run on the DGX host, which run on the switch, which require UFM. You know the command families: ibstat and rdma link show for physical layer, show interface counters and ethtool for traffic layer, show dcb pfc and show dcb ets for configuration layer.
You know the discipline: physical first, counters second, configuration third. You know the asymmetry between the NIC side and the switch side, and why show topology can reveal problems that rdma link show cannot.
And you know how the simulator works: one terminal tab per device, device-specific command sets, colour-coded prompts. The topology diagram is your map. Click a device, open a session, work that device's CLI. Exactly as you would in a real cluster.
The labs are waiting. In Lab 0, a GPU rail has gone dark. Click the DGX node in the topology to open your first session. You have everything you need to find the problem.