The U.S. Navy tried sodium reactors in the early days. In addition to Nautilus, the first nuclear submarine, they also built USS Seawolf, with a sodium-cooled reactor. Seawolf worked, but was enough of a headache that its reactor was replaced with a Nautilus-type pressurized water reactor. The US Navy didn't try sodium-cooled reactors again. Rickover on sodium-cooled reactors: "expensive to build, complex to operate, susceptible to prolonged shutdown as a result of even minor malfunctions, and difficult and time-consuming to repair."
The USSR tried sodium-cooled reactors, with moderate success.[1] They had leaks and fires too. One big reactor had 39 leaks over the years, none catastrophic. In the last 15 years of operation, none. Eventually they found all the construction defects.
Basic truth: complexity in the radioactive parts of a nuclear reactor causes trouble. Sodium reactors have fires. Pebble-bed reactors have pebble jams. (There's one in Germany that can't be unjammed or decommissioned.) Helium-cooled reactors leak helium. The nice thing about water-cooled reactors is that the complexity is on the outside, where repairs are possible. This is why new fancy reactor designs are viewed sceptically in the industry. You can't have any problems inside the reactor vessel, because you can't fix them.
[1] http://www-pub.iaea.org/mtcd/meetings/PDFplus/2009/cn176/cn1...