But by Thursday morning, the live video of the rig showed oil was still billowing from the top of the tap. A BP spokesman said the company had found a leak on the line attached to one of the valves, and that they were repairing it before moving forward. It was not clear how the discovery of the leak would affect the timing of the planned tests on the well. Thad W. Allen, the retired Coast Guard admiral who is leading the response, said that the test results would provide crucial information about the well’s condition, and “would determine our confidence to shut the well in and understand we’re not harming the well bore.” That knowledge would be particularly useful during a hurricane, he said, because it would prevent oil from leaking into the gulf after collection ships left for safe waters. From this Coast Guard cutter, it was apparent that the procedure to close off the well was starting. Flaring of oil and gas stopped about 3:30 p.m. local time on the two current collection ships at the well site, about a mile across the calm gulf waters. A gas- and oil-burning boom on one vessel, the Q4000, was extinguished first, followed by a gas flare from the Helix Producer, which created noticeable soot as it died out. The test had been delayed after government scientists raised concerns that the pressure buildup could damage the well. Earlier at the White House, Robert Gibbs, the press secretary, said that Energy Secretary Stephen Chu and others had been involved in the review, asking BP about the possible effect of the test on the well’s condition. Mr. Gibbs described the review as “a series of steps” that were being taken “in order to ensure that what we’re doing is being done out of an abundance of caution to do no harm.” Among the concerns was that if the well was damaged during the test, oil and gas might leak from the seafloor around the well rather than up through the well bore as it is now. Earlier, Kent Wells, a senior vice president for BP, said the test procedures were being reviewed to make sure scientists and engineers could interpret them properly. “We don’t want to end up with a test with inconclusive results,” Mr. Wells said in a morning briefing from Houston. He also announced that drilling of a relief well — considered the ultimate solution to stopping the gusher at its source — would be halted during the test as a precaution. The suspension of work would set the task back by a couple of days, he said. Earlier in the week, BP officials had said that if the test showed that the well could hold pressure, the valves might remain closed. That would end the gusher that began shortly after the Deepwater Horizon drilling rig exploded April 20, killing 11 workers. If the test shows the well is damaged, the flow of oil into the sea could still be stopped by increased collection of oil, a process that could continue for weeks, awaiting completion of a relief well. Admiral Allen said the government had asked BP for more information on the structural strength of the well. And in allowing the test to proceed, the government stipulated that pressure be allowed to build up in intervals, with acoustic tests to gauge the well’s condition every six hours. That would most likely lengthen the duration of the test, which had been expected to last from 6 to 48 hours. But the admiral said that the well would be monitored closely during the first three hours and that the test could be even shorter if the pressure stays low. The test brings to an end, at least temporarily, a period when more oil had been spewing from the well after a loose-fitting cap was removed to begin work on the new one. The old cap was diverting about 15,000 barrels of oil a day. If the pressure test shows that the well is damaged and the valves have to be reopened, full containment of the oil would probably not occur for several weeks, until other collection vessels could be brought in to handle more of the flow. That would raise total collection capacity to more than 60,000 barrels a day, the current high-end estimate of the well’s flow rate. Halting the gusher would then await the completion of the first relief well. The well site was a floating city on Wednesday, with scores of vessels scattered across the calm area. The activity was centered on the spot where the rig burned and then sank two days later. A huge drill rig, the Development Driller III, was about a mile away, working on the first relief well. A nearly identical rig, the Development Driller II, was drilling a backup that is not as far along. Supply boats attended both rigs, long sections of casing pipe on their decks. But the tongues of fire from the collection ships that had marked the site for weeks were out, at least for now.