NASA engineers drafted a nationwide team of troubleshooters Thursday to figure out why a candy-bar-size sensor malfunctioned in the external fuel tank of the space shuttle Discovery the day before, aborting the mission less than two hours before liftoff.

Deputy shuttle program manager Wayne Hale said "it is still theoretically possible" that planners could attempt a launch Sunday, "but this represents a really optimistic good-luck scenario which is not really realistic."

Hale said the experts -- hundreds of NASA scientists and outside contractors at agency centers nationwide -- were divided into 12 groups to develop a plan of action to be vetted by Discovery mission managers at Kennedy Space Center on Friday.

"At that point we'll make a decision about the way forward," Hale said.

The countdown for Discovery, which was slated to embark on the first shuttle mission since Columbia disintegrated on reentry 21/2 years ago, was called off Wednesday when a hydrogen fuel sensor deep in the external fuel tank failed to respond properly to test signals.

The scrub disappointed hundreds of dignitaries and well-wishers gathered to see the launch and left NASA grasping for answers, even as it strove to move past the lingering shock of the Columbia disaster on Feb. 1, 2003.

Discovery was standing on Launch Pad 39B on Thursday awaiting the managers' review and is still in a "scrub turnaround mode" anticipating the possibility of an imminent launch.

"But we're at a level of readiness that cannot be maintained," said Michael Wetmore, the space center's director of space shuttle processing. "When we get word from the troubleshooters, we'll either go forward [with the countdown] or back out."

Discovery is in the opening days of a launch window that extends to the end of July. A quick fix for the recalcitrant sensor could mean a fresh launch attempt within days, but extensive research, or returning Discovery to a hangar for repairs, could force a postponement until the next launch window, in early September.

"I wish I had more answers," Hale told reporters at a news conference.

Late Wednesday, engineers unloaded 1.6 million gallons of liquid oxygen and liquid hydrogen fuel from the orbiter, monitoring the hydrogen sensors throughout the process.

The four sensors in the bottom of the liquid hydrogen tank are about the size of small candy bars and are designed to trigger a main engine shutdown if the tank runs out of propellant before the shuttle gets into orbit.

The suspect sensor failed to respond to a test during countdown calling for a "dry" reading and locked instead in the "wet" position. "When the tank was drained [after the scrub], the sensor continued to show 'wet' three hours after it was dry," Hale said.

But then it switched to "dry" and began to respond properly. "This represents a problem for our troubleshooting team," Hale said, because engineers hoped to analyze the bad sensor in a "failure mode."

Hale said launch planners had sent out a call across the country, gathering the history of the external tank, the history of Discovery, original diagrams of the sensor system and the factory history of transistors in the system that were "not assembled to the best standard 20 years ago," when much of the shuttle technology was developed.

The sensor enigma originally arose in April when two hydrogen sensors cycled incorrectly during a fuel "tanking test," and extensive troubleshooting was unable to find the cause.

Engineers then swapped the 10-by-10-inch "black box" that serves as the system's controller and redid all the wiring. Even the sensors were different Wednesday, because Discovery had traded its April fuel tank for a newer one.

Still, the sensor malfunction remained an "unexplained anomaly" when project managers decided to try to launch anyway with the new parts. Although the decision may have appeared hurried in hindsight, several sources within and outside the agency said NASA had done almost everything it could to address the problem.

These sources, who spoke on the condition of anonymity because they are not in a position to speak for their organizations, said another tanking test would not have helped. Properly functioning sensors would have told engineers nothing, while malfunctioning sensors would have triggered an investigation that would have put a July launch in jeopardy -- no different than the current situation.

And the countdown test -- standard for every launch -- detected the anomaly. "You can't really fault them," said John Logsdon, a space expert at George Washington University. "It's easy in hindsight to single out something like this, but judgment calls like this are made all the time -- and basically the system worked. They didn't launch."

Discovery sits waiting on Launch Pad 39B.

Deputy shuttle program manager Wayne Hale, left, at a news conference, said "it is still theoretically possible" that planners could attempt a launch of the space shuttle Discovery on Sunday, "but this represents a really optimistic good-luck scenario which is not really realistic."