This is simply not true. A sustained nuclear chain reaction just needs to be physically reconfigured a bit to produce a reaction excursion. It's true that modern designs have made good progress in things like intervention-free failsafes that make such reconfigurations much less likely. It's simply not possible eliminate the possibility entirely.

And in any case, this is a piece of military hardware which might plausibly be "reconfigured" be being struck by a rocket. It's not a civilian reactor inside a secure facility.

Your statement is analogous to saying "charcoal grills do not start forest fires". Used correctly, they certainly shouldn't. But they sure as hell can if you kick them over on a dry summer day. Same deal here.

What the above poster is probably referring to are reactor designs that are passively safe. Like using lithium as a coolant and moderator. As the reactor heats up the lithium's density drops, which in turn reduces the rate of the nuclear reaction. It creates a stable equilibrium. The reaction would also stop if the lithium were drained - there would be no more moderator to sustain the nuclear reaction.

> reactor designs that are passively safe

To be glib: Not vs. fucking rockets they aren't. Those designs imagine that the fuel doesn't move and that the coolant is the medium of failure.

What happens when you move the fuel?

> What happens when you move the fuel?

It stops reacting. If you blow up a nuclear reactor you've made a nice radiological hazard (though Uranium/Thorium themselves aren't particularly hazardous), but as soon as the fuel stops being contained inside of the appropriate neutron-mediating environment and is scattered, it will become subcritical.

The question of whether or not deploying nuclear reactors in warzones is a good idea is entirely different from the question of whether meltdowns are a realistic possibility. A critical nuclear reaction can be sustained with neutron reflectors and moderators using an amount of fuel that would otherwise be subcritical. Any "reconfiguration" only reduces the criticality factor of such a reactor.

Even if a meltdown doesn't happen, if such a generator is struck by, for example, an artillery shell, who is going to clean up the mess?

Then the reaction stops because it's no longer in the presence of a moderator.