* Sodium Reactor Experiment (Leak, minor sodium explosion, decommissioned)[1]
* Monju Nuclear Power Plant (Sodium fire, never worked properly, decommissioned)[2]
There's even been a sodium fire at a solar plant, one of those big focused mirror systems.
Many of these new reactor designs are based on complex arguments that the worst-case accident doesn't require a huge, expensive secondary containment vessel capable of containing a major accident. That's a tough sell, since Chernobyl didn't have a containment vessel and Fukushima's reactors had ones that were too small. On the other hand, Three Mile Island had a big, strong containment vessel, and in that meltdown, it held, containing the problem. In all three accidents, the actual accident was worse than the design maximum credible accident.
The NRC is right to be skeptical of weak containment designs.
It's frustrating. The reactor designs that have worked reliably for long periods are very simple inside the radioactive portion of the system. Sodium reactors had leaks and fires. Pebble bed reactors had pebble jams. Helium gas-cooled reactors had leak problems. Molten salt reactors include a radioactive chemical plant. So nuclear power is stuck with water as a working fluid.