A full-blown war with North Korea would be "catastrophic," Defense Secretary Jim Mattis has said more than once.
A war, he says, would be "more serious in terms of human suffering than anything we've seen since 1953," the year the Korean War ended with an armistice, not a peace treaty.
In testimony before Congress last month, the veteran Marine commander and respected scholar of military strategy predicted victory for the U.S. and its South Korea and Japanese allies, but said, "It would be a war that, fundamentally, we don't want" and would win only at "great cost."
When fighting ended in the last war on July 27, 1953, almost 3 million people, military and civilian, had died on all sides, including 36,574 U.S. troops.
This time could be almost as bad, military experts say.
"To initiate military options takes you from a world of zero casualties to a world of tens of thousands in a best case, and quite possibly a million or more, especially if they use nuclear weapons," said Michael O'Hanlon, a senior foreign policy fellow at the Brookings Institution.
In other words, risking war with the unpredictable regime of Kim Jong Un is "unthinkable," except that for more than 60 years, an army of Pentagon planners has thought of little else.
Most of U.S. wargaming over the years has focused on the problem of repelling an invasion from the North.
The Pentagon's off-the-shelf war plan for the defense of South Korea, known as "OPLAN 5027" is updated regularly and includes a massive logistics scheme for quickly reinforcing the roughly 28,500 U.S. and 500,000 South Korea troops that would face off against North Korea's million-man army.
Only once since 1953 has the U.S. come close to launching an offensive, pre-emptive strike on the North.
In 1994, then-Defense Secretary William Perry ordered options drawn up for an attack on the North Korea's nascent nuclear reactor at Yongbyon.
The idea was to take out the nuclear facility with U.S. F-117 stealth fighters and cruise missiles before its fuel rods could be reprocessed into weapons-grade plutonium, estimated at the time to be enough to make a half-dozen bombs.
But Perry concluded the attack would likely incite the North to attack the South, and – weighing the same risks of all-out war that Mattis faces today – he pulled back.
"I decided against recommending that option to the president at that time," Perry wrote in his 1999 book, "Preventive Defense."
The risk calculus now is even more complicated, because North Korea has shown it can detonate a nuclear device underground and claims it can make a bomb small enough to fit atop a missile, a capability it has yet to demonstrate.
But the real deterrent to the use of military force by the U.S. remains an anachronistic, but fiendishly effective, array of artillery along the DMZ in range of the South Korean capital of Seoul.
"Depending on how you count, they have somewhere between 8,000 and 20,000 [pieces of artillery] within 30 miles of Seoul," says retired Air Force Gen. Chuck Wald, former deputy U.S. European commander.
"I don't know if it's clever, I hate North Korea," Wald told the Washington Examiner. "But it's pretty ingenious for them to put that many artillery pieces there. Because of the numbers, it's an overwhelming thing."
Many of the artillery pieces and multiple rocket launchers are dug into the sides of hills and cliffs to enable them to fire and then withdraw under cover.
Any war with North Korea would have to begin with a plan to take out those guns, which put teeth into Pyongyang's semi-regular threats to reduce Seoul to a "lake of fire."
It's those relatively low-tech conventional weapons, even more than the North's nuclear capability, that pose the biggest threat to Seoul's 20 million citizens and are the basis for Mattis' dire warnings about risks of all-out war on the peninsula. "It will involve the massive shelling of an ally's capital, which is one of the most densely packed cities on earth," Mattis told Congress.
Neutralizing that conventional threat would be a race against time, says professor Stephen Biddle, who teaches strategy at the George Washington University.
"This isn't going to be a sustained months and months-long campaign, because those artillery pieces won't live long enough for that," Biddle said. "The interesting question is: How many rounds can they get off before we destroy them?"
"I would overwhelm their artillery," said Wald, a former strategic planner for the Joint Chiefs of Staff. "I would put a plan in place to use every possible capability we have bomber-wise, and fighters for that matter, but mainly bombers to use the 30,000-pound bombs, which the B-2 and B-52s can carry two of those each."
Wald says the dug-in artillery positions are ideal targets for America's most powerful non-nuclear bombs: the 20,000-pound Massive Ordnance Air Blast, or MOAB, nicknamed the "Mother of All Bombs," which was dropped earlier this year in Afghanistan, and its bigger brother, the 30,000-pound MOP, or Massive Ordnance Penetrator.
"I would just pound the livin' bejesus out of the artillery until there is none left," Wald said. "They may get a few shots off, but if you do that at the right timing, there's a decision they have to make to start firing, and they may get a few minutes, but they are not going to get these hours where they are gonna fire on Seoul."
South Korea, well aware of the threat to its capital, has an extensive system of bomb shelters throughout Seoul, in theory enough for every resident to take cover during the early days.
In addition to neutralizing the artillery threat, protecting against nuclear attack would have to be the other top mission, suggests Brookings analyst Michael O'Hanlon.
Every available missile defense would have to be deployed including more Terminal High Altitude Area Defense batteries and ship-based Aegis anti-missile systems.
And then, the U.S. likely would try a decapitation strike on North Korea's capital to try to kill Kim Jong Un, or at least separate him from his military commanders, said O'Hanlon, who says even if that works it wouldn't end the war quickly enough to prevent a high number of civilian deaths.
"I think a shock-and-awe surgical strike still probably leads you into the tens of thousands of casualties; that's the absolute best case."
Once a war breaks out, the U.S. would have little choice but to go all in, Biddle argues.
"In all likelihood, if the U.S. military and the South Korean military decide they were willing to incur the cost, they could probably take Pyongyang," Biddle said. "Then, the issue becomes is there an insurgency that follows that? We took Baghdad in 2003. That didn't end the war."
And there's another "X" factor: China. It's unlikely Beijing, which sees the communist North as a buffer against the democratic South, would sit idly by and watch the war unfold.
"I think that China would probably enter the war, not to fight us, but to establish leverage in the north part of the peninsula," O'Hanlon said. "So, my expectation would be if there were large-scale conflict to break out, they would bring their own forces into at least the north part of North Korea, and if we were lucky they would tell us how exactly how far they were coming."
The risks of any war on the Korean peninsula seem to outweigh the benefits, but that would change the moment it appears North Korea might make good in its threat to target an American city with a nuclear-tipped missile.
"If the North Koreans were visibly about to launch a nuclear-armed ICBM at the continental United States, then I suspect the United States would decide it could tolerate a lot of South Korean casualties," Biddle said. "What we cannot do, in all likelihood, is engage in a major war in which we can guarantee that there are not thousands of civilian casualties in Seoul."
"I think that's what most of us mean when we say ‘no good military option,'" O'Hanlon said.