Hi Jelmer,
I think the issue is that your Pipe (TL) blocks are adding their own thermal mass and heat transfer, which is why the steady-state temperature changes when you modify pipe geometry. The pipes are being too "realistic" when you just want ideal conduction. Here is quick solution to your problem. For the 4 kW heat input connect your Controlled Heat Flow Rate Source to a Thermal Mass block first (set it to fluid_volume × density × specific_heat), then connect that to your pipe network via the thermal port (H). Now, for the 611 W/K radiator: Instead of using a pipe with convective transfer attached, connect a Convective Heat Transfer block directly between your pipe's thermal port (H) and a Controlled Temperature Source at 50 degrees centigrade. Set the heat transfer coefficient to 611 W/(m^2.K) with area = 1 m^2 to get exactly 611 W/K. For connecting pipes, I would make them thermally neutral by setting very short lengths (0.1 m), turning off "Dynamic compressibility" and "Fluid inertia", and setting the Nusselt number really low (0.001). Afterwards, you should see at steady state: 4000 W = 611 W/K × (T - 50 degrees centigrade), so T should stabilize at 56.55 degrees centigrade regardless of pipe geometry. In addition to that, I can even suggest simpler alternative, if you don't need fluid dynamics, skip Thermal Liquid entirely and just use regular Thermal blocks: a Thermal Mass for your fluid, a Heat Flow Rate Source (4 kW), and a Convective Heat Transfer block (611 W/K) to a 50 degrees centigrade source. This completely avoids the pipe complications.
The key is separating your thermal behavior (thermal mass + heat transfer) from the flow circuit (pipes just move fluid without participating in heat transfer).
Hope this helps!
