On the other hand, triggering the reactor to automatically shut down would be pretty easy -- and if you shut down all the nuclear reactors in the US for a few weeks, you'll certainly have made a significant impact.
Knock out several of those in critical locations, and you start to grind the U.S. economy to a halt.
You don't need to go nuclear. And destroying the engine was a fairly simple task of pushing it well outside its performance envelope.
EDIT: Looks like pnathan already cited this event -- see the last link in this comment:
http://news.ycombinator.com/item?id=3035909
It's a bit older than I remembered. The article is dated September, 2007.
I note incidentally that for one "catastrophic" scenario they describe, with an estimated "cost" of $700 billion, the damages figure now pales in comparison to what the U.S. economy has been through in the last few years. A bit of a lesson of its own regarding the rhetoric that surrounds the actual topic.
Of course there are safeguards to prevent catastrophes, but even stopping some part of the automation in an industrial plant could easily cause serious problems such as damaged equipment and downtime for debugging.
Computer-controlled systems that can have disastrous real-world consequences almost always have built-in checks to avoid these failure states.
In other words, even if you have an easily-exploited attack vector available to you, you still need to know a lot about your target in order to cause damage. Contrast this to guided missiles, which don't really need to know anything about the building that they are blowing up other than its GPS coordinates.
For this reason, I'm skeptical that "any dumb hacker" will ever be capable of causing something like a nuclear meltdown via virus infection.