Fast reactors have always had this potential problem, since they require a much higher density of high enrichment material. A larger fast reactor might have a ton of plutonium in it; in an accident where fuel is melting and moving around unpredictably it's hard to prove that a prompt supercritical mass of Pu won't assemble somewhere. Edward Teller famously warned about this in 1967.
This is one advantage of MSRs as fast reactors: the fuel is already dispersed in salt, so such rearrangement is less plausible.
- Not really. Check out the dramatic experiments at EBR-II (fast reactor) demonstrating unprotected loss of coolant at full power. It shut itself down. Without damage. And started back up later that day, and operated that year, as a research reactor, with better uptime (capacity factor) than the fleet at the time.
It would appear to carry an increased risk of hidden problems and groundwater poisoning. There might also be issues with corrosion and accessibility for maintenance.
That said, I'm not ruling it out. A lesson these new approaches to nuclear engineering is that we never figured out the right defaults the last time around. We were too focussed on making bombs and submarines.
> we already know the impacts of underground nuclear explosions
Nuclear plants should never fail by way of nuclear explosion. If this is a possibility it is a bad reactor design.