Rocket Lab found air pockets inside a rocket part that might never have failed in flight. The biggest rocket is delayed a year.
On 15 October 2026 Rocket Lab confirmed that Neutron's first flight has moved to no earlier than 2027 [1]. The company had switched the vehicle's first-stage propellant dome manufacturing from a third party's manual, hand-laid process to an automated fibre-placement machine. The goal was to remove the joints and inserts the old process left behind. It also let engineers thicken the forward and aft domes for extra strength. Inspection of the cured forward dome found voids: tiny air pockets, trapped where the thicker composite sections made it harder for heat and pressure to force resin through every layer [1]. Thicker sections cure less evenly than thin ones. Heat and pressure must drive resin through more layers before every trapped pocket of air can escape.
The voids, Sir Peter Beck said, "don't necessarily mean that the dome would fail in flight" [2]. He said the company will "proceed only when we believe the system has the appropriate margins" [2]. Beck put it more bluntly elsewhere: "Possible success isn't our standard" [1].
Engineers reproduced the defect in testing. They say the team was close to the correct manufacturing process on the first attempt. A replacement forward dome, built on a refined process, must still pass ultrasonic inspection before it counts as flight hardware. That step is targeted before the end of 2026. Full first-stage assembly and a pressurised, cryogenic test campaign are pushed into the first quarter of 2027 [1][3]. No new maiden-launch date has been named; Rocket Lab says it will share one once those milestones clear.
That is a year lost to a part that might well have flown. It is also the cleanest documented instance of fail-closed engineering this channel has had external material for this year. The slip is the margin decision working.
Composite manufacturing defects are common enough across the launch industry that one on its own rarely makes outside press at all. This one shows unusual discipline. Rocket Lab had a part that might well have flown. It chose not to find out, and said so before a single customer asked.
The regulator missed its own deadline too
Rocket Lab's decision would be the whole story most weeks. This week it sits beside two others that run the same shape through a different party.
The International Telecommunication Union's Space Services Department has more than 4,000 satellite network filings backed up awaiting review, with delays stretching to a year or more [4]. It is the regulator every satellite constellation depends on to publish its filing on schedule. Operators rely on that published record to show investors and lenders a credible path to deployment. A queue that long is a second clock an operator cannot wind forward itself.
Open Cosmos supplies the sharpest instance of what that backlog costs someone else. On 6 October 2026 the company asked the ITU's Radio Regulations Board for a one-year extension. The Board is due to consider the request at its November 2026 meeting [5]. Open Cosmos had missed fifty per cent deployment milestones in June and September 2026, across two filings covering 576 satellites at 1,050 kilometre altitude [5]. This series logged a related mechanism earlier this year: a different regulator's graduated, reversible demotion of a different operator's spectrum priority, conditioned on an outcome rather than enforced as a straight penalty. Open Cosmos's request is a second instance of that same shape, at a second regulator, with a different remedy. Two instances, two regulators. That is close to a pattern.
That backlog matters well beyond Open Cosmos. More than four thousand filings sit in one queue, a meaningful share of every active and planned satellite network on the ITU's books. Each one waits on the same administrative clock Open Cosmos is now asking to have paused. A regulator built to process a smaller industry has become the binding constraint for a much larger one.
This series has tracked a recurring pattern under the name Three Clocks: a company's manufacturing, launch and regulatory timelines run at different speeds, and only one is actually binding. This week, for the first time in the pattern's run, the slow clock belongs to the regulator.
Rocket Lab is held by its own cure cycle. Open Cosmos is held by a regulator's own filing backlog. The ITU itself administers an outcome-conditioned milestone regime, and it is missing its own processing schedule. Every operator in that four-thousand-filing queue is penalised by the administrator's throughput. In most prior instances of this pattern, an operator was held to a regulator's clock. Here the regulator is held to its own clock. The people waiting on it carry the cost.
That is worth sitting with for a moment. A deadline measured in satellites actually in orbit cannot be satisfied by paperwork, and a deadline measured in filings processed cannot be satisfied by better engineering on the applicant's side. Each test is honest about what it counts and unforgiving about what it ignores.
The Thursday series covers why a control design behaves this way, and what it costs when the binding layer moves from the party being measured to the party doing the measuring. This piece covers the three instances it produced this week.
None of this is a verdict on Rocket Lab's manufacturing competence or the ITU's administrative capacity beyond what the documented facts support. A composite dome that cures with voids in its thickened sections is a known failure mode in a known process change. The inspection built to catch it did exactly that. A four-thousand-filing backlog is a measured fact. It says the queue is long; it does not say why. Both read as documented engineering and administrative sequences.
The same company absorbed that year-long margin decision on Neutron. In the same week, it is the company New Zealand depends on to keep Electron flying on schedule from Māhia. The dome story and the Māhia story are two faces of one firm's manufacturing discipline.
The sensor New Zealand is about to host
New Zealand's anchor this week is a launch that, as at this writing, has not happened yet.
Rocket Lab's "We're Not Hawking Around" mission is scheduled no earlier than 21 October 2026, from Launch Complex 1 at Māhia. It carries three HAWK satellites, Cluster 16, for HawkEye 360 [6]. The three join a fifteen-satellite constellation spread across five orbital shells [6]. HawkEye 360 states its own constellation serves maritime domain awareness, GNSS interference detection, radar and spectrum monitoring, and defence and intelligence planning [7]. That is the operator's own published description of its purposes, not a claim this article makes on its behalf.
The launch site is New Zealand's. The satellites, the tasking authority and the data product are HawkEye 360's. Owning the ground beneath a launch pad is one capability. Owning the authority to direct what a sensor looks at next is another. The gap between them recurs across this series.
The Outer Space and High-altitude Activities Amendment Act 2025 is New Zealand's ground-based space infrastructure regime. It requires infrastructure supporting space objects to register, carrying an entity fine of NZ$250,000 and an individual fine of NZ$50,000. Its transitional authorisation period closed on 29 July 2026 [8]. That regime reaches ground infrastructure, launch and payload. It says nothing about who tasks a foreign-owned sensor once it reaches orbit. That is a different layer of sovereignty from the one the Act was built to govern.
New Zealand already carries a smaller, less visible version of the same dependency on the ground. The Southern Positioning Augmentation Network is a joint system with Australia, led on the New Zealand side by Toitū Te Whenua Land Information New Zealand. It sharpens GPS accuracy from five to ten metres down to as fine as ten centimetres for government and commercial users. It is due to move its satellite transponder service to a new Inmarsat I-8 satellite from 2027 [9]. A national positioning service depending on contracted capacity aboard somebody else's satellite is a smaller instance of the same pattern a radio-frequency geolocation launch raises at a larger scale: the capability a country hosts is not always the capability it directs.
No New Zealand government agency's own position on radio-frequency geolocation constellations as a sovereignty or maritime-surveillance asset is on the public record as at this writing. This article makes no claim about that absence; a future instalment will look at the gap.
What to watch
Four things carry real information between now and the next instalment of this series.
The ultrasonic inspection of Rocket Lab's replacement forward dome is the fact to watch, not any announcement of a new launch date. It is targeted before the end of 2026, and a clean pass there is the actual milestone. Rocket Lab has said a date will follow the first quarter 2027 assembly and test campaign, not precede it [1][3].
The ITU Radio Regulations Board's November 2026 consideration of Open Cosmos's extension request will say more about how that regulator actually administers its own backlog than this week's filing does on its own [5]. Check directly whether the "We're Not Hawking Around" mission flies on or near its no-earlier-than date, rather than assuming it; a launch window moves often in this series' own record. The Space Services Department's own four-thousand-filing backlog is a number to re-check in a few months. The queue does not clear itself.
Watch whether other operators follow Open Cosmos's lead and ask the ITU for relief of their own, rather than sitting in the four-thousand-filing queue without comment. A second and third request arriving in quick succession would say more about the backlog's real severity than this week's single instance does on its own.
The open-source dimension
The ground segment behind a radio-frequency geolocation mission runs on tools nobody has to buy from a single vendor. GNU Radio is the open-source software-defined radio framework used across amateur and professional ground stations alike. It lets an independent party process the same class of signal a constellation like HawkEye 360's own hardware is built to detect, without licensing a closed receiver chain from one company. That matters because an outcome-conditioned regulatory test is as good as somebody outside the operator being able to check the claim independently. This week's Rocket Lab and Open Cosmos material both turn on exactly that kind of test. Open ground-segment tooling is one of the few places that check can actually happen without an operator's cooperation. It hands into the sovereignty question this week's New Zealand launch raises next: who tasks the sensor, and who reads what the tasking authority chooses to share.
A radio-frequency geolocation satellite launched from Māhia for an allied operator puts that tasking sovereignty question on New Zealand soil. New Zealand hosts the launch site and, through its own allied partnerships, may see some of the resulting data. Owning the antenna and owning the authority to direct what it looks at next remain two different capabilities. New Zealand's ground-based space infrastructure registration regime governs who may operate ground infrastructure and launch from New Zealand soil. It says nothing about who tasks a foreign-owned sensor once that sensor reaches orbit. That gap between launch-site sovereignty and tasking sovereignty will matter more as more radio-frequency geolocation constellations choose Māhia.
A margin call and a filing backlog look like unrelated stories until you ask who gets to check either claim. Rocket Lab chose not to find out whether a flawed dome would have flown. The ITU's own backlog means thousands of operators cannot find out when their filing will clear. Open Cosmos asked for more time rather than waiting to find out what happens if it runs out. In every case, the party that could not simply assert its way past the question was the one that mattered. A replacement dome has to answer to ultrasonic data, not to a press release. A regulator's backlog has to answer to its own processing rate, not to the industry's patience.
What is the one dependency in your own organisation you have never tested? What would it take to find out before you need the answer?
This is one of seven weekly series in The Hamberger Report. Subscribe on LinkedIn and the next one arrives in your feed.
The views expressed in this article are entirely my own, informed by more than 30 years of professional experience in architecture, security, and technology leadership in New Zealand. I write as director of Te Pono Limited; the views are personal and do not represent the position of any client, any government agency, or the New Zealand government. My commentary on legislation and policy is analytical, drawing on publicly available sources and my professional expertise in architecture, security, and AI governance, and it is politically neutral.
About the Author: Andreas Hamberger is a New Zealand leader in Architecture & Security and Associate Member of the Institute of Directors. Space Mafia examines the sovereignty implications of orbital compute infrastructure.
This article was produced with AI assistance under my direction. Research, drafting and images pass through a pipeline I built and govern: automated gates for source verification, forbidden language and political neutrality, and my own review before anything is published. The tools include Claude, Gemini and Openart. The frameworks, arguments and editorial judgements are mine and are the same discipline I apply to the AI systems I audit for clients. AI accelerated the work; the thinking, and the responsibility for it, are mine.
[1] securities.io. "Rocket Lab Targets 2027 for Neutron First Launch After Dome Defect." 14 October 2026. https://www.securities.io/rocket-lab-targets-2027-for-neutron-first-launch-after-dome-defect/
[2] Starlust. "Rocket Lab Delays Next-Gen Neutron Rocket's Debut to 2027: Here's What Went Wrong." 15 October 2026. https://starlust.org/rocket-lab-delays-next-gen-neutron-rockets-debut-to-2027-heres-what-went-wrong/
[3] Fool.com. "Rocket Lab's Neutron Won't Launch Until 2027: Is the Stock a Buy 65% Below Its Record?" 15 October 2026. https://www.fool.com/investing/2026/10/15/rocket-lab-s-neutron-wont-launch-until-2027-is-the-stock-a-buy-65-below-its-record/
[4] Space Intel Report. "ITU: More Than 4,000 Satellite Network Filings Are Backed Up, Waiting for Review." 16 October 2026. https://www.spaceintelreport.com/itu-more-than-4000-satellite-network-filings-are-backed-up-waiting-for-review/
[5] SatNews. "Open Cosmos Seeks One-Year ITU Regulatory Extension After Missing Constellation Deployment Milestones." 6 October 2026. https://satnews.com/2026/10/06/open-cosmos-seeks-one-year-itu-regulatory-extension-after-missing-constellation-deployment-milestones/
[6] nextspaceflight.com. "We're Not Hawking Around, Mission Details." 2026. https://nextspaceflight.com/launches/details/6938/
[7] eoPortal. "HawkEye 360." 2026. https://www.eoportal.org/satellite-missions/hawkeye-360
[8] Ministry of Business, Innovation and Employment (New Zealand). "Our regulatory regime." https://www.mbie.govt.nz/science-and-technology/space/our-regulatory-regime
[9] Toitū Te Whenua Land Information New Zealand. "Accessing SouthPAN services." https://www.linz.govt.nz/products-services/geodetic/southpan/accessing-southpan-services