"The customer accepted the heritage qualification profile."
"Because we gave them the heritage profile."
Cameron's jaw tightened slightly.
"You're implying the customer was misled."
"I'm saying the profile is incomplete for this architecture."
"That's not the same claim you just made."
"It is technically the same problem."
Giselle spoke for the first time.
"Isaac, no one is asking you to compromise safety. We are asking whether the uncertainty in your model justifies stopping a program of this scale."
"I'm asking to test the uncertainty."
"At a four-month cost."
"At a four-month schedule cost."
She looked at him carefully.
"What do you believe happens if we fly the current design?"
Isaac could have softened the answer.
He did not.
"Eventually, we lose pressure authority. If the degradation reaches the range the model predicts during a critical maneuver, somebody will have to solve a failure on orbit that we can still solve on the ground today."
Michael Trent, the lawyer, wrote something down.
Cameron said, "The testing data does not validate that statement."
"The testing data does not recreate the condition."
"Then this is theoretical."
"All failures are theoretical before they happen."
After the meeting, Lyle found him near the elevator.
"How bad?"
"They want a narrower model."
"Meaning?"
"Meaning one that doesn't hurt the schedule."
Lyle looked back toward the conference room.
"Be careful."
Isaac frowned. "About what?"
"Being right in a way that costs powerful people money."
At the time, Isaac thought the warning was cynical.
Within three weeks, he understood it was descriptive.
Cameron's team requested revisions to the methodology section. The stated purpose was clarity. Several requested changes were reasonable. Others subtly moved the coupled-condition analysis into an appendix and emphasized that the model had not been reproduced independently.
Isaac made the changes that improved readability and rejected the ones that changed the conclusion.
Then he submitted the final version through Nexora's internal document system.
That version carried his name, the date March 14, 2017, and an explicit recommendation: do not sign flight approval until the coupled failure mode is resolved or independently disproven.
Three days later, Cameron called him into a smaller conference room.
No glass walls this time.
No engineers.
Michael Trent from legal was already seated.
Cameron placed a one-page certification form on the table.
"We need design sign-off to complete the milestone package."
Isaac read it.
"This certifies the propulsion assembly meets the mission environmental requirements."
"Based on the approved qualification standard."
"My report says the standard is incomplete."
"Your report says there's a theoretical coupled condition requiring further study."
"It says more than that."
"The customer has accepted the qualification basis."
"The customer hasn't seen my final analysis."
Michael Trent leaned forward.
"Isaac, the question here is whether you are refusing a normal technical certification based on analysis that has not passed peer validation."
"I'm refusing to sign something I don't believe is true."