Ten years on from its launch, the Hubble Space Telescope (HST) has proven to be an outstanding success.

In its orbit high above the Earth's turbulent atmosphere, the observatory has produced incredibly sharp and extraordinarily beautiful images of the cosmos.

But what was planned as the greatest scientific instrument ever built nearly turned into a $2bn white elephant.

A misplaced fleck of paint during the manufacture of Hubble's main mirror meant the first images sent down by the telescope were blurred.

It took some very clever thinking and an unprecedented rescue mission from the space shuttle to correct the fault.

As Hubble approaches its 10th birthday in April this year, the BBC Science programme Horizon has gone back to talk to some of the key players in the early drama - the team at precision optics company Perkin Elmer who were tasked with making the most perfect mirror ever built.

"The mission objectives were to observe objects 50 times fainter than we could do from the ground, and with 10 times better resolution," Hubble optical scientist Terry Facey tells the programme. "So to meet those goals the mirror had to be at least 10 times better than any previous mirror. We were going to strive for an almost perfect mirror, as close to perfection as we could possibly measure.

To make that mirror, Perkin Elmer brought together some of the best optical engineers in the world. Bud Rigby was head of the team that polished the mirror. Lou Montagnino was in charge of testing it.

They started with a huge chunk of glass that weighed nearly a tonne. First, they roughly ground it into the right shape. Then, a specially-designed computer-driven tool was used to polish the mirror to its precise concave shape, accurate to a millionth of an inch.

Each polishing cycle could last as long as two days. The whole mirror would then move to a special, vibration-proof chamber to test how its shape was progressing.

High in the testing chamber was an instrument called a null corrector, which sent calibrated beams of light from the mirror into a test instrument to check the developing shape. So sensitive was the system that the slightest disturbance would make testing impossible.

"All the testing had to be done at night," Bud Rigby recalls. "Cars as far away as three or four miles on the highway would still produce enough vibration in the test chamber that they had to stop testing."

The polishing and checking went on for a year. Lou Montagnino recalls: "A typical day was ten to twelve hours, it could run as high as sixteen or more. There were a few times when we could get out on a reasonable schedule, but Bud and I had to be available at all times."

Finally, some five years after the design process had begun, four years in storage waiting for the rest of the telescope to be built, and a further four-year delay because of the Challenger disaster, the mirror was finally sent into space as part of the completed observatory.

But, as everyone now knows, the mirror had a flaw called spherical aberration. It was slightly too flat, which meant that the light reflected from its edge, and light from its centre, were focussed in different places.

"It was shocking, I was upset by it," says Bud. "My association with the people that worked on that mirror, and that whole project, is very dear to me."

"There was an awful lot of time and energy put into this thing, a lot of personal commitment, a lot of technology, and to have it all be tarnished was very, very disappointing," adds Lou.

Unknown to anybody, a chip of paint had come off a measuring rod inside the null corrector. This exposed a chink of metal, and light hitting this chink distorted the measurements, producing an error that was copied onto the mirror.

"And that error was not picked up and resulted in the null lens being incorrectly built," says Chris Burrows from the Space Telescope Science Institute. "That error then got slavishly copied onto the primary mirror for the Hubble. And as a result the primary mirror was assumed to be correct because the null lens said it was, but in fact it was wrong."

The solution was ingenious: An instrument that would match the error in the mirror in reverse, and cancel it out. This optical 'fix' was called the Corrective Optics Space Telescope Axial Replacement (Costar). But it took an extraordianry mission from the space shuttle to go and fit it, and no-one was really sure it would work.

Horizon is broadcast on BBC Two on Thursdays at 2130 GMT