Read Chapter 3: FOUR MILLISECONDS

Four milliseconds became an argument that lasted three weeks.

Pike called it a non-repeatable anomaly. Bell called it instrumentation uncertainty. The project office called it a schedule risk. Luc called it a failure.

“Computers do not care whether a failure is repeatable,” he said during a design review. “If the power disappears once, the computer restarts once.”

A NASA representative named Harold Mercer rubbed his temples. “Mr. Moreau, are you recommending redesign?”

“I am recommending root-cause analysis.”

“That was not my question.”

“Then yes.”

Pike shifted in his chair. “We have one dropout in twenty-seven cycles. We repeated at temperature and could not reproduce. We have launch dates. We have integration dates. We cannot redesign every component that sneezes.”

Luc looked at the assembled men. “Then decide which sneeze you are willing to hear on the Moon.”

The room turned still.

Mercer ordered a limited investigation.

Luc and Reed opened unit 14 under a microscope. The suspect relay looked normal. Contacts clean. Coil resistance within tolerance. But when Luc heated the component and vibrated the board laterally, resistance across the contact jumped for fractions of a second.

“Contact bounce?” Reed asked.

“More than bounce.”

They tested three spare relays from the same lot. Two behaved. The third showed a similar transient.

Luc traced the purchase history. The relay had replaced the ceramic part under Engineering Change Order 1142. Reason: weight reduction and supply availability.

Qualification evidence: vendor certification plus Asterion vibration campaign.

The same campaign whose raw channels had gone missing.

Luc requested the vendor’s full qualification data.

Procurement sent twelve pages.

The document looked complete until he noticed the temperature range stopped at 65°C. DCP-7 qualification required operation to 85°C at the local board environment.

He took the file to Susan Bell.

“This relay was never tested to our required upper temperature.”

She scanned the pages. “There may be an addendum.”

“Do you have it?”

“No.”

“Did you have it when you signed the waiver?”

Bell’s face changed.

Not guilt. Fear.

“That waiver was approved by materials and project engineering.”

“Did you see the addendum?”

She lowered her voice. “Luc, leave it with me.”

He did not.

That evening Reed drove him to a diner near Webster where engineers ate under fluorescent lights and talked about the Moon over meatloaf.

“You scared Susan,” Reed said.

“She signed the document.”

“You think signatures mean people are free.”

“What do they mean here?”

“Sometimes that your boss was standing behind you.”

Luc stirred coffee. “That is not engineering.”

“No. That is employment.”

Reed told him Asterion had nearly lost a major guidance subcontract the previous year after cost overruns. Pike had been brought in to recover schedule. Since then, late tests had become passed tests, open issues had become watch items, and watch items had become footnotes.

“Has anything dangerous been hidden?” Luc asked.

Reed looked toward the counter. “Depends what you mean by hidden.”

“Daniel.”

“There was a battery disconnect on an unmanned drop article last summer. Power interruption. Guidance package shut down. Official cause was connector damage during handling.”

“What did you think?”

“I thought the connector looked fine.”

Luc’s hand stopped.

“Do you have the report?”

“No. But I know who wrote the first draft.”

“Who?”

“Frank Delaney.”

The same technician who had sworn through the glass when unit 14 failed.

The next morning Delaney did not come to work.

At noon, word spread that there had been an accident at the test annex.

A pressure vessel had ruptured during an overnight checkout. One man was dead.

Frank Delaney.

The four-millisecond event would have been easier to dismiss if Luc had not seen the computer emulator die.

He reconstructed the sequence with Reed frame by frame. Bus voltage sagged during switching. The regulator should have held output. Instead the suspect relay opened, the hold-up capacitor discharged, and logic supply crossed the reset threshold.

The reboot itself lasted less than a second.

In a laboratory, a second was trivial.

During lunar descent, a second could be geography.

Luc asked a guidance specialist what would happen if the computer reset during a critical phase.

The specialist frowned. “Depends on timing. Some states recover. Some need crew action. Some could cascade into abort logic.”

“Could it be catastrophic?”

“Anything can be catastrophic if you place it at the worst millisecond.”

That sentence bothered Luc more than reassurance would have.

He began running a simple probability exercise—not because he trusted the numbers, but because managers did. If the relay dropout occurred only once in thousands of switching events, perhaps mission exposure was small. But there was no valid failure rate because the part had never been fully qualified under combined heat and vibration. The absence of data was being treated as a favorable statistic.

At a review, Luc presented the logic.

“We cannot calculate reliability from tests we did not perform.”

A program analyst replied, “We have vendor heritage.”

“In what environment?”

“Industrial controls.”

“Factories do not launch on rockets.”

“Some factories vibrate.”

“Not through staging.”

Pike cut off the exchange.

Afterward, Harold Mercer found Luc in the hallway.

“You are technically right,” Mercer said.

Luc waited.

“That phrase usually precedes a request to become technically wrong.”

Mercer smiled despite himself. “No. It precedes context. NASA is trying to integrate millions of parts. We cannot demand bespoke proof for every resistor and relay.”

“I agree.”

“You do?”

“Yes. But criticality should determine proof. If a resistor fails and a lamp goes dark, accept heritage. If a relay failure can reset guidance, test the relay in the environment.”

Mercer studied him. “Write that down.”

Luc did.

The memo disappeared into a review chain and produced no immediate action.

Still, it mattered. For the first time he understood that change in a large program did not arrive like revelation. It accumulated in documents, objections, initials and repeated sentences until the organization could no longer claim not to have heard them.

He intended to keep repeating the sentence.