"Then don't give it to me."
"And let the satellite fall?"
"Your problem, not mine."
Eleanor studied him.
"You really will walk away."
"I already did once."
That was when she understood the most important thing about bringing Isaac Merritt back into aerospace.
The obstacle was not whether he still knew the work.
The obstacle was whether the institutions that had discarded him could behave well enough to deserve his help.
Six years earlier, Isaac had believed institutions corrected themselves when given enough evidence.
He was thirty-eight then, a senior propulsion-systems engineer at Nexora Aerospace in Houston, with twelve patents attached to his name and a reputation for making uncomfortable problems simpler rather than making simple problems impressive. He had joined Nexora because the company, at least in its recruiting materials, claimed to value that exact quality.
For the first several years, the claim seemed true. Isaac led failure reviews, redesigned a pressure manifold that had cracked under repeated thermal shock, and developed a diagnostic model later adopted across two propulsion programs. He was never the loudest engineer in a meeting, but when a room had exhausted its theories, people often looked toward him.
The Prometheus Drive Propulsion Assembly changed that.
The program was strategically important, technically ambitious, and late almost from the moment the contract was signed. Nexora had committed to deliver flight hardware to the Department of Defense on a schedule built partly around launch availability and partly around corporate promises made before several design questions were actually closed.
Those facts lived in separate parts of the organization. Program management owned the schedule. Finance owned the penalties. Engineering owned the hardware. Legal owned the contract language. No one, Isaac eventually realized, owned the consequences when all four collided.
The secondary pressure-regulation valve entered his attention through a routine test discrepancy.
A technician named Lyle Chen sent him a plot showing a slight response delay at the low end of the temperature band. The delay was tiny. It remained within the written tolerance. Most engineers would have signed it off as noise after confirming calibration.
Isaac asked for the raw data.
Then he asked for data from the prior three thermal cycles.
Then he asked for the vibration measurements from a different test team because one frequency band in the pressure trace bothered him.
Lyle had laughed over the phone.
"You know they're going to hate you for this."
"Who?"
"Everybody with a schedule."
"Good thing schedules don't vote in design reviews."
"You haven't met our schedules."
The first combined model took four days. Isaac ran it overnight on a workstation that hummed loudly enough to make his office wall vibrate. The output did not show a catastrophic failure. That would have been easier to explain.
Instead, it showed a tiny resonance condition appearing only when material stiffness shifted within a narrow thermal transition. The resonance lasted minutes, not hours. It did not exceed individual component limits. But each pass added a small amount of fatigue to the same region of the valve assembly.
The result was a failure mechanism made almost entirely from values that looked acceptable when reviewed separately.
Isaac requested a system-level review.
Cameron Ashford postponed it for six days because the program office was preparing a customer milestone presentation. During those six days, Isaac refined the model, widened the test assumptions, and became more certain rather than less.
By the time the review finally occurred, his presentation had forty-seven slides.
Eleven people sat around a glass conference table on the twelfth floor of Nexora's headquarters. Houston spread beneath the windows, flat and bright under spring haze. The room was cold enough that Isaac could hear the air-conditioning cycle during pauses.
Cameron sat at the far end. He was then Nexora's vice president of technical programs, a position that allowed him to speak for engineering when senior leadership needed certainty and for management when engineers needed permission.
Giselle Hartmann, the chief executive, sat to his right.
A legal representative named Michael Trent sat along the wall, which Isaac noticed because legal had never attended one of his technical design reviews before.
He began with the data, not the conclusion.
He showed the test discrepancy. He showed the thermal profile. He showed the vibration spectrum. Then he showed what happened when the two transfer functions were analyzed as a coupled system rather than separate qualification domains.
The predicted degradation curve crossed the mission safety margin between months eighteen and twenty-four, depending on actual orbital exposure.
By month twenty-eight, several scenarios showed loss of pressure authority severe enough to compromise controlled maneuvering.
Isaac did not use words like disaster or catastrophe. He did not need them.
"I am recommending a redesign of the secondary regulation assembly before flight qualification," he said. "At minimum, we need an integrated environmental test that reproduces the coupled thermal-vibration condition."
Cameron folded his hands.
"How much schedule?"
"For a proper redesign? About four months. Less if the housing modification works on the first iteration."
The room changed.
Four months was not an engineering duration in that meeting. It was a contract event.
A program manager named Deborah Lin looked toward Giselle.
"We don't have four months."
Isaac answered, "The valve doesn't know that."
Nobody laughed.
Cameron clicked back to an earlier slide.
"Your model includes operating conditions beyond the validated heritage range."
"It includes conditions the vehicle will experience."