The regulator says yes. Then the dates slip. The schedule that starts on the day of permission runs through proof, not capacity, and almost nobody puts that work on the plan.
The word "just"
On September 16, Kelly Ortberg, who runs Boeing, said a sentence I have been saying to myself all year. In May the Federal Aviation Administration had told him it was "absolutely comfortable with 42 to 47" aircraft a month. Four months on, he told an audience: "We're now driving at 47 a month, but we are not stable yet at 47 a month." And on the newest 737 variant: "We've completed all the flight testing. Now we're just in the documentation phase."
Just. The aircraft has flown. The engineering is done. What stands between Boeing and a delivered product is the assembly of proof, and the man running the company calls it the residual step. I think that in any operation a regulator watches, the residual step is usually the long one.
Two clocks
Permission is a regulator saying you may. Capability is doing it, again and again, at rate, with the records to show it. Everybody plans around the first clock. The second one starts the moment the first one stops, and it runs on evidence.
Before permission, the constraint is outside the building and everyone can see it. After permission it moves inside, to whatever work has the least visibility and the most hand-offs, and that is evidence work exactly. A machine has an uptime number. A supplier has a delivery date. The record that proves the line is stable has neither. It sits in pieces: a batch record in one system, a deviation in a second, a training log in a third, a supplier's test results in software nobody on the floor can log into. It starts late, because you only learn which proof is needed when an inspector asks. And it is serial. The reviewer cannot review until the author has assembled, and the author cannot assemble until three departments have answered their email.
There is a second reason it takes longer than planned. The procedure you wrote down and the way the work is actually done drift apart all the time, and the harder a plant runs to hit a new rate, the wider the gap. So producing the evidence is never printing what exists. It is reconciling two versions of a Tuesday. Inspectors know this. Practitioners who reviewed seven months of inspections under the FDA's revised quality rule describe inspectors who evaluate how information flows across living processes rather than checking whether the documents were renamed. A flow that was never recorded has to be reconstructed, and reconstruction is the schedule.
Three months and four filing cabinets
Stellantis's own chronology for a 2026 recall, filed with the safety regulator, records an investigation into rearview camera failures opening on May 7. Through May and June its safety and engineering groups pulled warranty data, field records, production records and customer assistance records. On August 6 a committee determined that 844,027 vehicles did not comply. Three months from question to answer, and the chronology names the reason itself: four kinds of record, held by four groups, had to be brought together before anyone could decide.
The contrast came the same month. Stryker recalled nearly 12,000 defibrillators and told customers what to do in the same breath as telling them what was wrong, because its statement rested on its own internal testing. Where the evidence is produced as the work happens, the gap between knowing and acting is days. Where it has to be gathered afterwards, it is months.
I will add our own, because the argument should hold for the company making it. In September our engineering team put a rebuilt core of our product into production one week late. The software had been built for weeks. What stood between built and shippable was a validation run, and the internal description of the target was "proper validation, not the make-believe validation we were doing earlier." That is the second clock, running in my own building.
The case that it is only physical
The best objection: Boeing's instability is suppliers, parts and people, and calling it an evidence problem flatters a software point of view. Documentation is short next to years of design. Three months to diagnose an intermittent fault across a million vehicles may just be what careful engineering takes.
Mostly true. Some of the gap after permission is physical and always will be. My claim is narrower and, I think, harder to escape: the evidence part of the gap is the part nobody scheduled, nobody owns and nobody measures, and that is what puts it on the critical path. A physical constraint gets a capacity plan. An evidence constraint gets discovered when someone asks for the proof. And if documentation really were the trivial last mile, companies would forecast it accurately. Across aircraft, vehicles and devices, they do not.
So: start a second clock on the day of permission instead of stopping the first one. Give evidence work what physical work already has, an owner, a date and a measured duration. Capture decisions and approvals as they happen so the record is a byproduct of the work rather than a project that follows it. GE Aerospace is helping suppliers build safety management systems the regulator does not yet require of them, on the reasoning that "between the shop floor and the runway lies a chain of decisions, processes, and handoffs." And notice that some permissions come with their own clock: ISO 9001:2026 was published on September 16, and certificates to the old edition expire on September 30, 2029. Three years is roughly how long it takes a plant to notice its procedure and its practice have drifted apart.
How I lose this bet
If someone measures post-permission schedules properly and the evidence work turns out to be a small, stable slice with physical capacity doing all the delaying, I was wrong, and I will say so here.
The cleaner test is comparative. If companies that capture their records as the work happens get from permission to stable capability no faster than companies that reconstruct them afterwards, the mechanism I have described does not exist.
And one with a date. On September 16 Boeing said certification of the 737-10 was coming "very soon." If that aircraft is certified and the line is declared stable at rate before March 31, 2027, then the documentation phase was the short part this time, and I will put that at the top of this post rather than the bottom.
