Andy Garcia · freelance dev
FRENES
Back to the blog
Space

Starship Lost One of Six Engines and Still Reached Orbit. The Go Rule Was About the Way Back

October 20, 2026 7 min read

On Monday, September 28, 2026, at 7:48 a.m. Central time, Starship lifted off from Starbase, Texas, on its 14th flight. The 13 flights before it had all been deliberately suborbital. This one was SpaceX's first attempt to put the ship in orbit, with 26 Starlink V3 satellites in its payload bay.

Seven minutes and 18 seconds into the flight, by NASASpaceflight's log, one of the ship's six Raptor engines shut down. A few minutes later, SpaceX's Dan Huot told webcast viewers: "We did have one of those engines go out on the way uphill, and because of that, we are not going to do the additional burn that will send us into orbit." Then he paused while controllers deliberated on loops the webcast did not air. "So, we are not giving up yet on orbit." About ten minutes later: "the go/no-go poll for orbit has finished, and we are go for orbit." (The quotes are as transcribed by The Space Review.)

One engine fired for 19 seconds. Starship was in orbit, and the satellites went out.

The reversal was the drama of the day. What caught my attention is a sentence SpaceX had already published on its mission page by September 15: "Starship will only execute a burn to enter orbit after the flight control team has ensured there is sufficient redundancy on hardware critical to doing the subsequent deorbit burn at the end of the mission."

I don't build rockets. I'm CTO of a LegalTech built on Claude, I ship iOS apps and a strategy game, and I orchestrate a good part of my work with Claude Code. That sentence answers a question every builder meets sooner or later: when part of your system has just failed, what decides whether you take the step you cannot undo?

The safe path came first

SpaceX's explainer "Starship to Orbit," dated September 15, describes the design. On its first 13 flights, the upper stage flew a "passively safe" suborbital trajectory: when the engines stopped, it lacked the speed for orbit, so "even if control over the spacecraft was completely lost, it was already on a path to reenter the atmosphere and splash down in a pre-determined location."

Flight 14 kept that default. The ascent would still end on the same suborbital path toward the Indian Ocean. Controllers would check the vehicle during a coast, then decide: orbit, or let it come down. Orbit required a second burn, by a single sea-level Raptor.

Every burn after the climb was planned on sea-level engines: one for the orbit insertion, one for the deorbit, three for the landing. The capability had been rehearsed. On Flight 13, on July 24, the ship relit a single Raptor in space, which SpaceX called "a core capability for future orbital missions." The ship's three vacuum-optimized engines had no role after the climb, as Huot would explain on air.

How small the step to orbit is

Orbit is about sideways speed more than altitude. SpaceX's explainer puts it at about 17,500 mph in low Earth orbit, roughly 7.8 kilometers per second: fast enough that the Earth curves away beneath the vehicle as fast as it falls.

After engine cutoff, NASASpaceflight logged Starship on a path of -40 by 278 kilometers. The 278 is the high point. The -40 means that the lowest point of that orbit, if the Earth were not in the way, would sit 40 kilometers below the surface: the path meets the planet before it completes a lap. The insertion burn, by the same account, added 95 meters per second and left the ship in a 262 by 275 kilometer orbit.

So 95 m/s, about 1.2 percent of orbital speed, raised the low point by roughly 300 kilometers, from a trajectory that ends in the ocean to one that circles the Earth. That is why this burn is the gate. Before it, the system ends somewhere safe if nobody does anything. After it, a ship that cannot relight an engine has no controlled way down, and the way down is exactly the hardware the rule named.

Fourteen minutes from no-go to ignition

The sequence, from SpaceX's pages and NASASpaceflight's timestamps:

  • T+7:18: the vacuum Raptor shuts down. SpaceX says the five remaining engines burned "for longer than planned to compensate for the loss of thrust" and still reached the planned suborbital trajectory. Cutoff came at T+8:35, about 24 seconds after the time in SpaceX's pre-flight timeline.
  • T+11:20 (T+11:30 in Astronomy magazine's log): the webcast announces there will be no orbit.
  • About a minute later: the data is still under review.
  • After a wait of nearly ten minutes (Astronomy magazine logged it at T+21:53): the poll comes back go.
  • T+25:20: the insertion burn starts, three seconds after the T+25:17 in SpaceX's final pre-flight timeline (the September 15 version had T+25:28).

SpaceX's summary gives the basis: the team "determined the three sea-level Raptors engines needed for subsequent burns were healthy."

The safe trajectory is what made those 14 minutes affordable. While the team deliberated, the ship stayed on a path that already ended in a pre-determined splashdown zone, and the burn still went out on schedule.

Then they came home early

The satellites left the bay between T+35:05 and T+1:04:22, and SpaceX says it made contact with all 26. The plan called for about six orbits over a flight of nearly 10 hours, with splashdown west of Chile.

Instead, in SpaceX's words: "Out of an abundance of caution given the engine issue experienced during ascent, the flight control team decided to limit the duration spent on orbit." Starship headed for one of two landing zones coordinated before launch and cleared of sea and air traffic. One sea-level Raptor performed the first Starship deorbit burn; all three relit for the landing burn and flip. Splashdown north of Hawaii came at T+3:08:30, after just under two orbits.

The booster had its own engine losses. One of its 33 Raptors shut down on the climb, 31 of the planned 33 relit for the boostback, and 11 of the planned 13 for the landing burn, before a splashdown in the Gulf. SpaceX then triggered the booster's flight termination system on purpose, as a demonstration of the safety hardware.

What the evidence does not show

The cause. SpaceX's flight page, read on October 20, gives none. NASASpaceflight reported on October 18 that Elon Musk, in an interview, described a software failure isolated to that engine's computer, a "glitch." That is a characterization, not a published investigation. The same outlet counts this as the second vacuum Raptor lost on ascent in three flights of the current vehicle generation.

The deliberation. Huot was relaying decisions made on loops the webcast did not air. Why the first call was no-go, and who argued what, is not public. I will not guess.

Reliability. A single flight, with engines lost on both stages, does not establish it, and Flight 14 did not fly the ten-hour profile. As of October 20, SpaceX's target for its next flight was November, with an attempt to catch both stages, according to NASASpaceflight's October 18 report. Musk said in September that his main concern was the ship staying intact through reentry, since a return to Starbase means reentering over land.

What I take from it

  1. Write the go rule before launch, and aim it at the way back. SpaceX's explainer stated the general condition, "assuming the vehicle is healthy." The mission page made it precise: it names the hardware the return depends on. For a deploy or a data migration, as an analogy, the equivalent question is whether the rollback path is still intact.
  2. Make the default end safe. A passively safe trajectory makes inaction a valid outcome. In software terms, loosely: the risky step needs an explicit yes, and doing nothing leaves the system where it was.
  3. Check what the next step needs, not whether the dashboard is green. A failed component that nothing downstream uses is information, not a veto.
  4. Expect the first call to change, and make changing it cheap. The public heard no-go, then go. The safe default gave the team room for that: about ten minutes of review, with the burn still on time.
  5. Bank the objective, then shrink the exposure. Once the satellites were out, the team took a pre-cleared early exit instead of flying nearly seven more hours after an engine anomaly on the climb.

Starship went to orbit with one engine out because the rule SpaceX had published 13 days earlier asked about the way home, and the way home was intact.

Sources

A project like this one?

I design and deploy products like this. Let's talk.

Let's talk