Of all the things that could go wrong in developing a missile defense system, Pentagon officials figured that the booster would be among the least of their worries.

They figured wrong.

Despite decades of U.S. experience building rockets to launch satellites, space shuttles and other craft -- and despite the involvement of such industry giants as Boeing Co. and Lockheed Martin Corp. -- the booster project veered far off course. Its persistent problems have led to a two-year hiatus in intercept tests and driven the Pentagon to fall back on an alternate booster design.

The troubled booster program illustrates the difficulties the Pentagon has confronted in pursuing a workable missile defense system -- and doing so within the administration's compressed timetable of deploying by the end of 2004. Interviews with many who have worked on the project or have closely observed it point to a long history of miscalculations.

Projections were overly optimistic, technical challenges were underestimated, designs were faulty, and workmanship was poor. Repeatedly over the past few years, program officials have declared the corner turned, only to acknowledge months later, with renewed frustration, that problems have persisted.

"The booster story typifies the risks we run when we think we know too much, when we're very comfortable with something," said Air Force Lt. Gen. Ronald T. Kadish, who recently left the Missile Defense Agency (MDA) after five years as director. "We didn't pay attention to the risks involved in that part of the program."

The booster's role is to carry a "kill vehicle," a 120-pound package of sensors, computers and thrusters. Once in space, the kill vehicle separates from the booster and speeds toward an enemy warhead, destroying it in a high-speed collision.

In producing the booster, Pentagon officials had hoped at the outset to save money and accelerate development by relying on commercially proven motors assembled largely from off-the-shelf components.

"Back in 1999, we thought off-the-shelf technology was kind of a slam-dunk," Kadish said in an interview.

But the missile defense mission presented some particular design challenges. In contrast with satellite shots, where the trajectories are known, the booster in a missile intercept mission must be prepared for a wide variety of flight paths, ranging from relatively flat to very lofted. The flatter trajectories, Kadish said, "require more thrust control, and that became more of a challenge than most designers appreciated."

Additionally, Pentagon officials wanted the interceptors packaged in canisters to make transporting and handling easier. That affected performance, though, and complicated maintenance and repairs.

Boeing, which already had responsibility for overseeing the entire missile defense system, won the contract in 1998 for producing the booster. Quality-control and design problems began to emerge early.

One especially nettlesome problem arose from a requirement that the interceptors might have to stay in the ground for years before being fired. This precluded use of hydraulic systems to control the boosters' nozzles, because such systems are susceptible to leakage and malfunction.

So the only option was a battery-powered system. But none had ever been produced for such a powerful booster.

Other problems followed. In the spring of 2000, technicians discovered a weakness in the system for controlling the roll, pitch and yaw of the booster, and they struggled with a redesign. They also found heat damage in a metal line that piped hot gases to the rocket's thrusters. The line had to be re-engineered.

Still another setback resulted when the writing of computer code used to guide the kill vehicle turned out to be more cumbersome than expected. That added more weeks to the schedule.

By then, Kadish and other senior MDA officials had lost confidence in Boeing's management team. The company shook up its missile defense group, replacing the executive in charge and others. But problems continued into 2001 -- among them, poor bonding in adhesives for some booster components.

A test launch in August 2001 succeeded, but a second test in December 2001 flopped. Immediately after launch, the rocket tumbled end over end, spewing gas and fire out its front end as well as its rear. An investigation showed that the dome of the motor had been cracked when the booster was moved to the launch site.

With the booster program by then 11/2 years behind schedule, Pentagon officials approved a plan to re-bid the project. To increase the chance of getting an effective booster, they decided in early 2002 to authorize two separate development efforts based on two different designs.

One contract went to Orbital Sciences Corp., which had a good record supplying NASA and the Pentagon with small rockets, target missiles and lightweight satellites. Orbital's proposal promised a faster, cheaper booster. But the company had never built a large, three-stage missile -- the kind the Pentagon was seeking for the missile defense system.

The second contract went to continue Boeing's design -- but not with Boeing in charge. The project was taken over by Lockheed Martin, which had more experience building boosters and had already supplied boosters for early flight tests.

"It's not unheard of, but it is unusual," said Lockheed's Doug Graham, recalling the decision to have his company take on the task of completing Boeing's design.

Said Jim Evatt, a Boeing senior vice president: "We did run into normal developmental and technical issues on our booster. But if Lockheed had not wanted to build it, we would have continued with the program."

Lockheed made some changes in Boeing's design. It added a cold-gas attitude-control system to reduce the booster's tendency to roll after launch. It also integrated an avionics package into the booster, doing away with the glue pads that had attached the package to the dome and were prone to come loose. Additionally, the company reconfigured the booster's flight computer.

But Lockheed officials put off other desired improvements, intent on meeting the Pentagon's deadline. The transition from Boeing had already taken 15 months, about half a year longer than expected.

"There were other things they wanted to do, but, quite frankly, cost and schedule wouldn't allow them," said Col. Damian Bianca, manager of the interceptor program for the MDA.

By last fall, program officials were still hopeful the booster could be ready for deployment this year. Then disaster struck.

A couple of fluke explosions occurred at a propellant-mixing complex in San Jose belonging to Pratt & Whitney's Chemical Systems Division, supplier of the upper-stage motors for Lockheed's booster. The second blast, in September 2003 -- triggered when a contractor mistakenly cut a pressurized line -- killed the man and caused Pratt & Whitney to suspend mixing operations.

Lockheed, driven to look for another motor supplier, lost any chance of having its booster ready in time to satisfy President Bush's 2004 deployment date. Meeting that deadline would depend on Orbital, the other booster supplier.

Founded in 1982 by three friends not long out of Harvard Business School, Orbital had concentrated on research and development, not soon-to-be-deployed weapons systems. With several hundred million dollars in annual revenue, the company was still a bit player in an industry characterized by a few big-name firms that earned over $20 billion a year.

Orbital seized on the booster project as a ticket to the major leagues -- and also a way out of debt. Like almost all U.S. satellite builders and rocket companies in the 1990s, Orbital had gone deeply into debt investing in commercial satellite networks.

Pentagon officials decided to take a chance with the small, financially burdened firm because of its record of solid performance. They also liked the fact that its design would be based on the company's Pegasus rocket, which had failed just three times in 31 launches of small satellites.

"Was there concern going with a smaller company? A little bit, but not overwhelming," recalled Maj. Gen. John Holly, manager of what the MDA calls the Ground-based Midcourse Defense system. "They were not as big as Lockheed but they had a proven track record."

In early 2002, the company was given one year to produce a prototype booster for a test launch. The launch, which took place 13 months later, was a success.

"I think that really broke the ice," said J.R. Thompson, Orbital's president. "We changed from being a stalking horse to being a real contender to provide the booster."

But converting Pegasus into a missile defense booster wasn't simple. Pegasus is launched from underneath an airplane and has wings and control fins to help it fly. Orbital had to strip off the wings and fins and figure out how to shoot it from a silo.

"There were two ways missiles got out of silos in the U.S. missile experience," said David W. Thompson, one of Orbital's founders and currently its chief executive. One was by using the rocket's own thrust. The other was by injecting a compressed gas into the system before igniting the rocket engines.

"We couldn't do either one of those," he said. "Instead, we needed a system that could actively steer the vehicle as it flew out of the silo, and the tolerances were pretty tight. We described it as taking the Enterprise out of space dock at warp power, so you had to be pretty careful about how you did that."

The missile defense mission presented Orbital with additional challenges. The guidance and navigation components were complex. So was the software used to link the interceptor to the system's command and control network.

Further, space launches normally allowed plenty of time to get ready. With Pegasus, for instance, the countdown tended to span a couple of days. But with missile defense, the booster would need to blast off within minutes. That meant a design that could allow for both long-term storage in a silo and short-notice activation.

"I think we went into it with our eyes pretty wide open and had made a pretty realistic assessment of the technical challenges in front of us," said Ronald J. Grabe, a former astronaut who runs Orbital's launch vehicles business unit.

One approach Orbital considered but rejected was canisterizing the booster as the original Boeing -- now Lockheed -- design had done. While the canister made it easier to move the booster, it also complicated maintenance. Doing without a canister, Grabe said, "seemed to us to be a little more straightforward."

The initial demonstration launch in February 2003 marked the first time Orbital had included a steering nozzle on the first stage of its Pegasus rocket. The launch also achieved successful separation of the first and second stages.

"With that flight, we addressed about 60 to 70 percent of what we thought were the high-risk items," David Thompson said.

A second launch in August 2003 demonstrated many remaining items. And a third launch in January this year was the first to fly with the full deployment configuration, using a mock kill vehicle. By then, Pentagon officials were talking to Orbital about speeding up production.

Originally, the Pentagon had planned to use a mix of Orbital and Lockheed boosters in the initial deployment of 16 interceptors at Fort Greely in Alaska and four at Vandenberg Air Force Base in California through 2005. With Lockheed sidelined until 2006, Orbital has expanded production to meet the full demand for boosters in this first batch.

The booster project has become Orbital's single biggest program, representing just under 30 percent of company's total revenue last year and employing 25 percent of its workforce.

But the Orbital booster has yet to be flight-tested with an actual kill vehicle. The first such test, due months ago, has been postponed repeatedly as a result of technical issues with the booster as well as with the kill vehicle, which is made by Raytheon Co.