Days later, it was obvious even from a distance that something had gone terribly wrong at SpaceX’s Cape Canaveral launch pad.

NASA brought reporters and photographers to a mound near Space Launch Complex 41 on the morning of Sept. 7 to observe the rollout of an Atlas 5 rocket that would launch the agency’s OSIRIS-REx asteroid sample return mission the next day. But their attention was drawn instead to the scene a couple miles away at SLC-40, SpaceX’s pad.

The top of one of the lightning towers at the pad, normally white, was blackened. The strongback, the equipment that cradles the rocket as it’s moved horizontally to the pad and then raised to the vertical position, was still standing at the pad, but also blackened. Its top section was sagging and trailing cables or wires.

That damage was the result of an accident at the pad on the morning of Sept. 1 that destroyed the Falcon 9 standing there and its payload, the Amos-6 communications satellite. While any particular launch accident is surprising, this one was especially true, as it took place not during the launch itself but during fueling of the rocket for a static-fire test two days before the scheduled launch.

The explosion could be felt for miles, including at neighboring SLC-41, where the Atlas 5 and OSIRIS-REx were being prepared for launch. “People felt it. It really rattled our building, shook them up a bit,” United Launch Alliance Chief Executive Tory Bruno said in a Sept. 8 interview.

Bruno said they had to evacuate the pad for a while, forcing them to play catch-up for a couple days to keep the Atlas 5 launch on schedule.  But other than some “special cleaning” of the rocket afterwards to ensure no contamination of the spacecraft, he said the accident had no effect on what was ultimately a successful launch. “The rocket and the spacecraft were well-protected and they were fine.”

More than a week later, the accident remains a mystery. SpaceX’s last formal statement was Sept. 2, when it said that the explosion originated around the upper stage liquid oxygen tank, leaving open whether it originated with the rocket itself or with ground support equipment.

In a series of tweets in the early morning hours of Sept. 9, SpaceX Chief Executive Elon Musk said that investigators had made little progress in determining the cause of the accident. “Turning out to be the most difficult and complex failure we have ever had in 14 years,” he wrote.

He added that the accident was particularly mysterious since it took place during a “routine filing operation” about eight minutes before the first stage engines were scheduled to briefly ignite in the test. “Engines were not on and there was no apparent heat source.”

Whatever the cause, the accident leaves SpaceX and its customers, big and small, in limbo. The accident came as SpaceX appeared to be hitting its stride: the company had performed five launches in the previous four months for NASA and several commercial customers, almost all launching on the very first attempt. Just two days before the accident, SES announced that it would launch its SES-10 satellite in October on the first reflight of a Falcon 9 first stage, the latest milestone in SpaceX’s efforts at reusability.

SpaceX needed to maintain that launch cadence because of a busy schedule. Counting Amos-6, SpaceX planned to carry out as many as nine more Falcon 9 flights through the end of the year, mostly for commercial customers as well as another NASA cargo mission to the ISS. Those launches are now on hold indefinitely as SpaceX investigates the accident and fixes whatever went wrong.

For now, customers both big and small appear willing to be patient. “They’ve revolutionized the launch industry. Our engineers made the right choice in 2009,” said Thomas J. Fitzpatrick, chief financial offer of Iridium, at an investor conference Sept. 8. He was referring to the decision Iridium made to launch its next-generation constellation on the Falcon 9 even before that vehicle’s first flight.

Iridium was up next after Amos-6, launching the first 10 satellites on a Falcon 9 from Vandenberg Air Force Base in California in the second half of September. Six more Iridium launches were to follow by the end of 2017. Now the company must wait as gaps begin to appear in its original constellation of satellites, launched in the late 1990s.

“Yes, [SpaceX] have had a mishap. Our confidence in them is not shaken,” Fitzgerald said. “We’re sure they are going to figure out what happened and get back in business. They really are an economical provider.”

NASA, heavily invested in SpaceX for commercial cargo and crew missions to the ISS, is also sticking with the company. “If you’ve been in this business long enough, you’ve experienced a day like last Thursday,” said Tim Dunn of NASA’s Launch Services Program Sept. 6. “We look forward to their return, soon.”

At the other end of the customer spectrum is The Planetary Society, the space advocacy organization that is flying LightSail-2, a cubesat designed to test solar sail technologies, as a secondary payload on the second Falcon Heavy mission. That launch, last scheduled for the spring of 2017 after considerable delays, will likely be pushed back again.

Bill Nye, the chief executive of the society, said his confidence in SpaceX wasn’t shaken by the failure. “Those people are going to pick themselves up and figure this out,” he told reporters at the Kennedy Space Center Sept. 8. “Falcon Heavy? Bring it on.”

Few alternatives

That confidence and bravado may sound like an endorsement of SpaceX, but it also reflects that, for many customers, they have little choice but to wait and hope the Falcon 9 returns to flight soon. With Proton launches delayed by an anomaly during a June launch, Orbital ATK’s Antares yet to return to flight from a launch failure nearly two years ago, and Arianespace’s manifest filled for the foreseeable future, there are few opportunities for SpaceX customers to switch to another vehicle if they tire of delays.

ULA’s Bruno sees the current market as an opportunity for Atlas, which traditionally has not been price competitive with other commercial vehicles but does offer a strong record of success. “There’s been a lot of interest in Atlas,” he said. “I have slots reserved and available for people next year and the year after.”

There’s also the complication that, even if Falcon 9 is able to return to flight relatively soon — it took SpaceX six months to start launching after last June’s Falcon 9 failure — repairing the pad at SLC-40 may take much longer.

SpaceX hasn’t disclosed the extent of the damage to the pad, but it’s clear from photos that much of the infrastructure at the pad will have to be repaired or replaced. When the Antares crashed and exploded next to its launch pad at Wallops Island, Virginia, in 2014, the closest recent analog to this accident, repairs to the pad took nearly one year and cost $15 million.

That won’t affect some customers, like Iridium, launching from Vandenberg. And SpaceX says Launch Complex 39A, the former Apollo and shuttle-era pad at Kennedy Space Center it’s refurbishing for commercial crew and Falcon Heavy missions, could also be used for Falcon 9 satellite launches. The pad, the company said, should be operational in November.

But while the alternative pad might be operational in November, it’s not clear when Falcon 9 will be operational again. If the investigation drags on for weeks or months, the industry’s confidence in SpaceX might start to become as damaged as the launch pad caught in a sudden, unexpected fireball.