The hook. In today's space digest, there was a line that an engineer cannot pass by, but an ordinary reader will: "End of the Cluster mission: Samba and Tango will burn up in the atmosphere." Right below it — another: "Ariane 6 launched MTG-I2 — Europe completed the meteorological trio." Two news items, one day, one silence in both cases. And I thought: here's a perfect picture of how European spaceflight ends eras — quietly, meticulously, in a remote corner of the Pacific Ocean, at 11 kilometers altitude in a business jet cabin, with three dozen synchronized cameras aimed at the sky a second before a four-ton spacecraft turns into a ball of plasma.
While SpaceX does standup about Mars and celebrates every failed Starship test as "successful data collection," ESA has been sending a Falcon 900 (a Dassault business jet, not Elon's rocket — important not to confuse them — legacy of French aviation) for two years in a row to catch its retired satellites in the sky. September 8, 2024 — Salsa. October 2025 — Rumba. August 31, 2026 — Samba. September 1, 2026 — Tango. Four identical spacecraft, two years, two continents, and the final seconds of each one recorded on camera from an aircraft flying over the South Pacific Ocean Uninhabited Area — the largest spacecraft graveyard on the planet.
And here's what really got me: nobody is discussing this. This isn't "Crew Dragon saved the crew," not "Starship reached India," not "Starliner stuck on ISS." This is a European engineering finale where everything goes strictly on schedule, nobody takes selfies, nobody tweets "just did it," and in thirty years this episode won't make it to Netflix. But that's exactly why it's worth telling.
Investigation.
To understand why four satellites on August 31 and September 1, 2026 must burn up in the atmosphere over the South Pacific, we need to rewind history thirty years.
In 1982, the European Space Agency approved the Cluster mission — four identical satellites that were supposed to fly in a tetrahedral formation (regular triangular pyramid) in a highly elliptical polar orbit with perigee at 19,000 km and apogee at 119,000 km. The idea was revolutionary: before Cluster, all magnetospheric missions flew solo. And the magnetosphere isn't an object, it's a process: solar wind hits Earth's magnetic field, the field bends, current sheets emerge, shock waves, reconnections of field lines, auroras. To understand how this works in three dimensions, you need four spacecraft, spaced apart, measuring the same thing from different points simultaneously.
Earth's magnetosphere is a giant drop: it starts 65,000 km from the planet on the day side and stretches 6,300,000 km on the night side. It's an environment where plasma physics works in pure form, without laboratory walls. Solar wind (a stream of charged particles from the Sun, ~400 km/s in quiet times) collides with it and forms a bow shock (bow shock wave) — an analogue of a sonic boom, but in plasma. Behind the bow shock lies the magnetopause, where solar wind pressure balances Earth's magnetic field pressure. Even deeper — the magnetotail, which stretches millions of kilometers and where magnetic reconnections occur — the main process Cluster was created for.
Each of the four spacecraft carried the same set of 11 instruments: FGM (magnetometer), EDI (electron drift instrument), ASPOC (potential control), STAFF (field fluctuation analyzer), EFW (electric field), DWP (digital wave processor), WHISPER (electron density sounder), WBD (wideband receiver), PEACE (electron spectrometer), CIS (ion spectrometer), RAPID (high-energy particle detector). 71 kg of scientific payload on each. Spacecraft mass — 1200 kg, of which 650 kg was fuel for orbit corrections. Diameter 2.9 m, height 1.3 m, solar panels at 224 W, spin rate 15 revolutions per minute.
The distance between spacecraft could be changed from 3 to 60,000 km depending on the scientific task. When the satellites are close — they "dig into" the fine-scale structure of the magnetic field in near-Earth space. When far — they get a panoramic picture of large-scale processes. Cluster flew both inside and outside the magnetosphere, crossing its boundaries every 57 hours. This was the first mission in history capable of measuring the magnetosphere in three dimensions.
Launches were planned for 1995–1996 — on Ariane 5, a new European launcher that was supposed to replace the proven Ariane 4.
June 4, 1996, 12:34 UTC, from the Kourou spaceport in French Guiana, the first test flight of Ariane 5 launched — the rocket on which ESA and CNES had bet everything. Flight V88 (as it's officially called, by Arianespace's internal numbering) lasted 37 seconds.
At the 37th second of flight, the inertial navigation computer (SRI — Système de Référence Inertielle) saw a horizontal velocity value that didn't fit into a 16-bit signed integer. Ada code inherited from Ariane 4 without changes (for "commonality," as stated in the commission report) tried to convert a 64-bit float to a 16-bit signed int — and got arithmetic overflow. The exception handler for this was disabled "for efficiency reasons." On Ariane 4, this code never ran in flight — the realignment function was used only before launch, and was supposed to shut off in the air, but by inertia continued working for 40 more seconds after launch. On Ariane 4, within 40 seconds, horizontal velocity didn't manage to reach a value breaking the 16-bit representation. On Ariane 5 — acceleration was five times higher, and within 40 seconds, horizontal velocity exceeded the allowable range.
Both SRI computers — primary and backup — failed simultaneously (they ran identical code). At the 37th second, the rocket lost both sources of inertial data and switched to emergency mode with diagnostic values that the development team didn't even bother to connect to nozzle control. The nozzles received a command for "large correction of deviation that didn't exist," the rocket sharply changed course, aerodynamic forces tore the hull apart at the 39th second. The automatic destruct system worked normally. Ground cameras caught a cloud of debris over Guiana.
Payload: four Cluster satellites FM1–FM4. Damage — $370 million (just for the spacecraft, not including rocket cost). This was the first Ariane 5 flight, and it lasted 39 seconds.
The Inquiry Board report (published July 1996) became a classic case in software engineering. The main conclusion wasn't about arithmetic overflow — but about lack of analysis of what variable values could actually occur on the new trajectory. The development team assumed that realignment code on Ariane 5 doesn't work in flight — but didn't verify this assumption. Pre-flight tests didn't include Ariane 5 trajectory data. A function needed only for ground preparation went into flight by inertia, and nobody thought to disable it. This wasn't a bug — it was systemic failure of thinking.
Code analysis of Ariane 5 after the disaster became the first example in history of large-scale static analysis using abstract interpretation — work performed under Patrick Cousot's direction at the INRIA Paris laboratory laid the foundations of modern critical software verification tools (Frama-C, Astrée, Polyspace).
When Ariane 501 exploded, ESA had one spare spacecraft — Phoenix (FM5), an identical copy built in case any of the four FM1–FM4 needed replacement during operations. After the disaster, the decision was made quickly: Phoenix would become one of the four Cluster II spacecraft, and three others would be built anew based on Phoenix with minimal changes. And since Ariane 5 hadn't flown successfully yet, Russian Soyuz with Fregat upper stage was chosen for Cluster II launch — through STARSEM (European-Russian joint venture promoting Soyuz on the Western market).
The industrial contract went to Dornier (Germany). Phoenix's payload was returned to providers for restoration — some electronic components stopped being manufactured in four years, limited redesign was required. ESA covered 40% of instrument expenses from its budget, the rest through national agencies of participating countries.
Key political moment: ESA had to ask the scientific community to "move over" other missions in the queue, because Cluster II was taking resources. Solidarity of scientific communities is a separate topic, but important: implementing Cluster II meant delays for other projects, and the community accepted them because they understood the uniqueness of the data.
Two launches — two months:
The names of the four satellites carried the aroma of an era when ESA loved giving missions "dance" names: Rumba, Salsa, Samba, Tango. Four Latin American dances in polar orbit — counterintuitive, but quite in ESA tradition: Solar Orbiter, BepiColombo, Juice, Rosetta — all get names either mythological or cultural, never military.
And here arose the second problem: Soyuz's first stage performed with reduced thrust, and the Rumba+Tango pair ended up in the wrong orbit. They had to be pulled to working orbit at the expense of their own propulsion system and Fregat upper stage thrust. Philippe Escoubet, ESA Cluster Project Scientist, recalled twenty years later: "ESA was a bit worried 20 years ago, during the launch of the second pair of spacecraft. Ever since then, the mission has made huge progress, and it is far from finished."
That was in 2000. The mission was designed for two years.
Cluster outlived its stated service life by 12 times. From 2000 to the end of 2022 (formal mission end), four spacecraft worked more than 22 years, and until the final weeks of 2026 continued transmitting data from elliptical polar orbit. This is rare even by standards of long-lived space missions — most spacecraft die from fuel depletion or solar panel degradation long before formal completion. Cluster survived because its architecture was originally conceived as repairable in orbit (as far as possible without astronauts): propulsion systems with fuel margin, instrument redundancy, ability to change distance between spacecraft.
Main scientific results of Cluster (according to "Cluster After 20 Years of Operations" review, AGU, 2021):
Magnetic reconnection in 3D. The mission's main hit. Magnetic reconnection is a process where magnetic field lines of different polarities "stitch together" anew, releasing enormous energy. This process controls solar flares (which can be a billion times more powerful than an atomic bomb), magnetic storms on Earth, plasma behavior in fusion reactors. Cluster gave the first 3D image in history of the central point (magnetic null point) in the reconnection zone. Before Cluster, theory said: the null point must exist. Cluster showed what it looks like — the magnetic field in this zone is twisted into a tube 500 km wide. This became possible only thanks to tetrahedral architecture: one spacecraft sees the local picture, four — assemble it into three-dimensional reconstruction.
Black auroras. Cluster confirmed that black auroras — strange dark spots inside polar lights — are "anti-auroras", regions that pull electrons from the ionosphere rather than pumping them from above. This discovery was impossible with one spacecraft: only simultaneous measurements by four probes at different altitudes showed that in these zones the electric field points down, not up.
Magnetic vortices the size of Earth. Cluster discovered that under certain conditions, solar wind forms magnetic whirlpools larger than Earth that "bite into" the magnetosphere and pump charged particles into it. This explained the mechanism of solar wind penetration inside the magnetic shield — something that was a mystery before Cluster.
Bow shock and foreshock. Cluster studied the magnetosphere's bow shock wave and the region before it (foreshock), where solar wind is "preliminarily disturbed" by reflected particles. This data is critical for space weather forecasting.
Space weather and satellite protection. The main practical result: Cluster gave basic models of how magnetic storms damage electronics on satellites and in ground power systems. Without Cluster, we still wouldn't know how many hours before a solar flare we need to put satellites in safe mode and why exactly some of them fail while others don't.
When Cluster celebrated its 20th anniversary on August 8, 2020, ESA published an article with the headline "Cluster's 20 years of studying Earth's magnetosphere." Philippe Escoubet said then: "Cluster observations have uncovered details about the processes in the magnetosphere, revealed how the atmosphere supports life, and provided essential insights into space weather."
This was a quiet revolution. Cluster never made the covers. Cluster didn't take beautiful pictures of galaxies. Cluster photographed Earth's magnetosphere in 3D — and this turned out to be more important than all the beautiful pictures combined.
By the early 2020s, fuel on the four spacecraft was running low, and solar panels were degrading. ESA had two options:
Wait for natural orbital decay. Cluster in highly elliptical orbit isn't ISS. Its orbit is determined by Earth, Moon, and Sun gravity, and predicting the atmospheric entry point is extremely difficult. They could go for uncontrolled reentry over an arbitrary point on the planet — with risk to population.
Do targeted reentry — controlled descent over a pre-selected area with minimal risk.
ESA chose the second path. The target was South Pacific Ocean Uninhabited Area — a giant stretch of ocean between New Zealand and South America that the UN formally designated as a "safe splashdown zone" for spacecraft. This is also where stages of Russian Protons and other launchers go. This is the largest spacecraft graveyard on the planet.
For each of the four spacecraft, ESA conducted a series of maneuvers in the final weeks of life to direct them into a specified entry corridor. Precision — on the order of tens of kilometers. This isn't "hitting the window" like a crewed capsule, but it's a thousand times more precise than natural descent.
Salsa — September 8, 2024, 18:47:36 UTC ± 4 seconds. First targeted reentry of a scientific satellite in history. Burned up over South Pacific Ocean Uninhabited Area, south of Easter Island. Observation was conducted from a Falcon 900 (Dassault business jet, chartered through Astros Solutions). 26 synchronized cameras at six instrument stations. Infrared observation time — 23 seconds. Stijn Lemmens, Senior Space Debris Mitigation Analyst at ESA: "With better data on exactly when and how they heat up, break up, and which materials survive, engineers can design satellites that burn up completely — so-called design-for-demise satellites." And a separate surprise: the spacecraft continued transmitting telemetry after passing the 100-kilometer mark at a speed exceeding 10 km/s. Nobody expected communications to hold that long. Infrared camera data showed that atmospheric density differed from models by 20% along the trajectory, and structural breakup began slightly earlier than models predicted.
Rumba — October 22, 2025, also over the South Pacific. Without airborne observation campaign (for budgetary reasons or because Salsa had already answered the main questions).
Samba — August 31, 2026, 23:41:54 CEST, uncertainty ±10 minutes. From a Falcon 900 that flew from an airport on Tonga island (the westernmost accessible point to intercept the entry corridor).
Tango — September 1, 2026, 23:33:20 CEST, also ±10 minutes. Again Falcon 900, again from Tonga, again three hours flight to observation point.
Between these two reentries, the observation team will return to Tonga, refuel, rest, and fly out again the next night. Three hours flight each way to catch 23 seconds of phenomenon.
And here begins the most interesting part, which is actually why I dug into this topic.
Stijn Lemmens told Aerospace Global News (February 12, 2026) details that ESA usually doesn't disclose in official releases. Falcon 900 (Dassault business jet, not to be confused with SpaceX rocket — this is a different Falcon, legacy of French aviation) — is a three-engine long-range business jet normally used for VIP transport. Price — about $50 million in standard configuration. It has range 7400 km, ceiling 15,500 m, cruising speed ~500 knots (926 km/h).
The problem: a spacecraft enters the atmosphere at ~11 km/s (almost 7 miles per second, as Aerospace Global News writes). Falcon 900 flies 40 times slower. The aircraft cannot "catch up" with the satellite. So the strategy is different: the aircraft positions itself in advance under the predicted trajectory and waits. When the spacecraft enters the atmosphere, it first moves toward the aircraft, then flies past. This is a fly-by in the literal sense — the operator manually tracks the object through optical and infrared cameras until it goes over the horizon.
To maximize observation time, the crew performs coordinated turns during the pass. The turn point determines which segment of the trajectory gets captured. Too far along the trajectory — you'll miss initial heating. Too far behind — you'll miss final fragmentation. These maneuvers are rehearsed in advance, and in the air one team member calls out current angles while camera operators adjust instruments based on attitude sensor readings. Twenty-three seconds — and that's it.
The aircraft can't fly anywhere: airspace restrictions and safety zones dictate where exactly it can be. And the most subtle thing — the aircraft must take off from a point, reach the entry corridor, conduct observation, return for refueling, rest, and repeat all this for the second spacecraft in 24 hours. To make this possible, ESA conducted fine orbit corrections of Samba and Tango on January 19 and 20, 2026: one reentry shifted slightly east, the other slightly west, so both events remain reachable from one airfield.
Beatriz Jilete, space debris systems engineer at ESA: "Moving two satellites to meet a plane sounds extreme, but the unique reentry data we'll collect is worth orchestrating the challenging encounter over a remote stretch of ocean."
The main unknown result of the two final reentries: will Samba and Tango be able to transmit data deeper into the atmosphere than their predecessors. Bruno Sousa, Cluster operations manager at ESA, explained why this is possible: Salsa and Rumba went into safe mode during the last perigee pass due to solar panel overheating. Samba and Tango's panels degraded less. If they stay active until the last perigee, ESA could get temperature telemetry from inside the spacecraft at depths down to 110 km — where no aircraft flies and an inspector satellite cannot be (it would burn up with the object).
This would be a unique combination: internal data from the spacecraft itself + external optical and infrared data from the aircraft. Linking "what engineers see from outside" with "what the object feels from inside." If successful — Cluster will give its last scientific result not on the magnetosphere, but on reentry physics.
And this is no longer about Cluster. This is about Draco — ESA's next mission, launch scheduled for 2027. Draco is a controlled kamikaze satellite with more than 200 sensors, four cameras, and a protected capsule to preserve data after breakup. The idea: Draco will enter the atmosphere itself and film its own destruction from inside. What's impossible to see from outside because the object is already burning. Stijn Lemmens: "With three practice runs under their belt, the team will be able to link the observations made from the plane to what's happening within Draco at exactly that time."
So Salsa, Rumba, Samba, Tango are three rehearsals for Draco. Each Cluster reentry is a rehearsal for the next mission. And these rehearsals give what no laboratory can give: real data on exactly how a spacecraft breaks up in a real atmosphere. Wind tunnels and arc-jet facilities reproduce extreme temperatures, but not real trajectory, not real air density, not real material composition. And Cluster gives all this with accuracy to several seconds.
The final goal — "design-for-demise" satellites: spacecraft designed to burn up completely upon atmospheric entry. Not a single piece reaches the ground. This requirement is already introduced in ESA regulation for new missions, and Cluster is the only program in history that actually tests the models underlying this requirement.
Here we need to step out of engineering optics for a minute and look politically.
In a world where SpaceX celebrates every Starship launch as "paradigm shift," and China puts its own weather satellites in orbit, ESA does what it does best: finishes the mission. No fanfare. No press tours to Cape Canaveral. No memes.
Cluster II started with the failure of 1996. Ariane 501 — disgrace, investigation, $370 million damage. ESA could have closed the program, as NASA would have after such a thing. Instead, the agency took the spare Phoenix spacecraft, reassembled Cluster II in four years, and launched it on Soyuz. Because Cluster isn't just a mission, it's knowledge infrastructure about the magnetosphere needed for space weather forecasting, for satellite protection, for future crewed flights to the Moon and Mars.
Today, when we look at Artemis II and future Mars missions, we assume astronauts will be protected from solar radiation. This protection relies on models built on Cluster data. Without these models, flight to Mars is Russian roulette with a dosimeter.
And the Cluster finale — targeted reentry with observation from an aircraft — isn't just "how to dispose of a satellite." It's how to turn the end of a mission into a source of new knowledge. Each of the four spacecraft dies in the atmosphere, but before that it works: its sensors continue to measure, its structure continues to be tested, its behavior continues to surprise engineers. And this data forms the foundation of Draco — a mission that will fly specifically to burn up and tell us what happens inside a spacecraft at the moment of death.
This is the spaceflight I like. Not media-friendly, not viral, not Twitter-worthy. Quietly working machines that faithfully send data for twenty-five years, then honestly fall into the atmosphere on schedule, simultaneously turning their death into the next mission.
A couple of curious details that didn't fit into the main story but make the picture fuller.
Cluster and IT disasters. Ariane 501 is a classic case in software engineering, alongside Therac-25 and Boeing 737 MAX. Every senior software engineer should know this story: 64-bit → 16-bit conversion without checking, disabled exception handler, shared code between two systems with different physical characteristics. If you work with critical software and don't know about Ariane 501 — that's a red flag.
"Phoenix" as metaphor. ESA named the spare spacecraft Phoenix — after the mythical bird that rises from ashes. After Ariane 501's failure, Phoenix truly became the "resurrected" fourth part of Cluster II. This is the only time ESA gave a satellite a name not connected to mythology or culture — but the justification was too obvious to refuse.
Salsa reentry and "photo of the day." On September 9, 2024, ESA published the first image of Salsa's reentry — a small bright dot in the upper part of the frame, shot through infrared filter from a Falcon 900. This is one of the rare photographs of controlled scientific satellite reentry in real time. The shot was taken by Ranjith Ravichandran and Gerard Armstrong from the Astros Solutions team.
Falcon 900 crew commander during Salsa observation — a specially certified pilot trained for observation campaigns. This is a rare specialization: most civilian Falcon 900s fly with VIP passengers. Observing a spacecraft on reentry is a mission pilots aren't prepared for in normal life.
Space weather and economy. According to ESA estimates, damage from a Carrington Event-level (1859) solar storm to the modern world economy would be $1–2 trillion in the first 24 hours. Cluster is one of the main data sources for models that allow predicting such storms hours in advance. That is, a mission that cost ~$1 billion (including Ariane 501 and all extensions) protects infrastructure worth twenty thousand times more. ROI of space science isn't an abstract figure, it's concrete insurance against a concrete cataclysm.
Irony of the finale. Four Cluster satellites flew in pairs on two Soyuz rockets in 2000. Russian-European cooperation through STARSEM was the norm then. Today, in August 2026, ESA looks at Cluster's finale as a technological triumph, while ESA–Roscosmos relations are on pause due to the war in Ukraine. One of the few Soyuz rockets still flying for European missions is the one that will launch Draco in 2027, if the contract isn't frozen.
It's always seemed to me that the main beauty of spaceflight is in finales. Not in launches, not in press conferences, not in "Mars here we come" tweets. But in how a spacecraft dies when its work is done.
Cluster lived 25 years. That's longer than the career of the engineer who launched it. Cluster saw the first 3D image of magnetic reconnection. Cluster discovered black auroras. Cluster found magnetic vortices larger than Earth. Cluster survived after Ariane 501's failure, after the abnormal orbit of the second pair, after solar panel degradation. And now Cluster is dying the way it lived: quietly, meticulously, on schedule, transmitting data to the last second.
Its death is not a disaster, but an experiment. Four identical spacecraft enter the atmosphere under four different conditions, and each reentry gives engineers data for the next mission. This isn't a metaphor. This is literal: Salsa, Rumba, Samba, Tango are three rehearsal runs for Draco, which will fly in 2027 to burn up and tell us what happens inside a spacecraft at the moment of death.
And here's what I took from this digging. When SpaceX celebrates every failed test as "great progress" and promises Mars flight in five years (the eighth time in a row), ESA quietly for two years in a row sends a Falcon 900 business jet over the South Pacific Ocean Uninhabited Area to catch a burning fragment in the sky lasting 23 seconds. Not for a press release. Not for a tweet. So that the next satellite burns up completely, and not a single piece kills a person on the ground.
There's something very old and very European in this. Pedantry as a form of heroism. And silence as a form of respect for the work.
Cluster will end on September 1, 2026 at 23:33 CEST. Nobody will notice. But I noticed. And now you have too.