Orbital data published on 17 October 2026 put Cosmos 2612 about 1.5 km from ICEYE-X36 and closing. ICEYE-X36 is a Finnish-American radar satellite in a constellation Ukraine tasks for intelligence. The question of what a collision cascade would look like with this many satellites in low orbit usually gets an answer built on accidents. This approach falls outside the published cascade models, which measure accidents.
Russia launched six Cosmos satellites on one Soyuz in April. In May four of them raised their inclination by 0.8 degrees, spending about 105 m/s of delta-v to match the orbital plane of ICEYE-X36 at roughly 547 km, according to Greg Gillinger of Integrity ISR. Plane changes cost so much fuel in low orbit that operators rarely make them. Gillinger calls plane matching the first and most expensive step toward rendezvous and proximity operations. ICEYE-X36 weighs about 90 kg, and he assesses it lacks the fuel to respond to a co-orbital threat.
Moscow named this class of target in advance. At the UN First Committee in October 2022, Konstantin Vorontsov of the Russian foreign ministry said Western commercial satellites helping Ukraine were "quasi-civilian infrastructure" that could become legitimate targets for retaliation. The same report noted that Ukrainian volunteers had bought access to ICEYE imagery and full use of one ICEYE satellite.
The Bystander Problem
Russia has also tested the hardware. In July 2020 US Space Command said Cosmos 2543 released an object near another Russian satellite in a non-destructive test of a space-based anti-satellite weapon. Jonathan McDowell measured the release at about 700 km/h relative to its parent. In November 2021 a Russian missile destroyed Cosmos 1408 between 460 and 490 km and produced more than 1,500 trackable fragments plus hundreds of thousands of smaller pieces. The crew of the ISS, flying about 60 km below, took shelter in their spacecraft.
The bystander problem in orbit dates to 1962. On 9 July 1962 the United States detonated the 1.4 megaton Starfish Prime warhead about 400 km above Johnston Island. A 2010 Defense Threat Reduction Agency report traces the early loss of several satellites in the following months to the radiation belt the blast created. The losses included Telstar 1, the first satellite to relay live television, and Ariel 1, Britain's first satellite. The cost of the 1962 test fell on Britain and on AT&T, parties outside the decision.
Washington now says Russia is developing a nuclear weapon for an orbit far more crowded than the one Starfish hit. In February 2022 Russia launched Cosmos 2553 into an isolated orbit near 2,000 km, in a radiation band other satellites avoid, and US officials link it to a nuclear anti-satellite program. In April 2024 Russia vetoed a Security Council resolution reaffirming the Outer Space Treaty ban on nuclear weapons in orbit, with 13 members in favor and China abstaining. The next week John Plumb, assistant secretary of defense for space policy, told Congress the weapon would treat every satellite in its reach the same, whatever its owner, and could leave low orbit unusable for up to a year. In March 2026 General Stephen Whiting of US Space Command told the Senate a detonation would put every country's satellites in low orbit at risk, Russia's and China's included, and in April his command ran its first Apollo Insight wargame on the weapon. Moscow denies the program and calls Cosmos 2553 a research satellite, which by 2025 appeared to be tumbling.
Models Built on Accidents
A megaconstellation is a swarm. Thousands of narrow units each carry their own avoidance authority and coordinate through a shared catalog, the shape Brief №7 drew from Arquilla and Ronfeldt. Cosmos 2610 to 2613 have the geometry of a small hostile swarm inside it: four satellites matched to one target's plane, which lets some track while others close, the functional subsetting Brief №4 described. Russia started building that geometry at launch, within half a degree of the target's ascending node, and completed it with the May plane change.
The cascade work published in 2025 and 2026 measures accidents. The CRASH Clock estimates the time until a catastrophic collision if every operator stopped maneuvering at once, and its authors name a major solar storm or a catastrophic software fault as the cause. The value fell from 164 days in 2018 to 2.5 days in May 2026, and the authors set the danger zone below 1.4 days. Hugh Lewis and Donald Kessler found the band from 800 to 1,000 km already above the runaway threshold in 2025. In July 2026 Lewis found six of 16 planned constellations large enough to drive runaway debris growth on their own. Their stability model counts the fragments that random collisions produce.
A nuclear detonation supplies the CRASH Clock's premise by design. Satellites disabled by radiation stop maneuvering as they fail, and they stay in their shells as uncontrolled objects. The condition the clock measures then holds for every operator in reach, Russia included.
Collision avoidance works as a cooperative protocol among thousands of operators who answer to separate authorities. Starlink alone made 207,152 avoidance maneuvers between December 2025 and May 2026. Each one depends on knowing where the other object will be. In December 2025 SpaceX blamed a near miss with a CAS Space satellite on operators who keep their ephemeris to themselves. A hostile satellite keeps its plans secret and steers toward its target. Conjunction screening treats it as one more object whose orbit data goes stale with every burn.
The Sequence at 547 km
A strike or a released object at 547 km would start the sequence. Fragments spread above and below the break, as Cosmos 1408 showed when its debris reached the station 60 km lower. In January 2026 SpaceX announced it would move every Starlink at about 550 km down to 480 km during the year, because debris and planned constellations are sparser below 500 km. Fragments from 547 km decay through that new shell on their way down. Every tracked fragment generates conjunction warnings and avoidance burns, and a COMSPOC study found each burn degrades orbit forecasts for days, with position errors up to 40 km. The CRASH Clock authors estimate that cataloging half the debris from a collision takes about 100 days. For those months operators dodge objects they track poorly and add position error for everyone around them. An attacker can produce the loss of situational awareness that the CRASH Clock assigns to solar storms.
Altitude limits the damage at 547 km. Drag pulls debris down within years at that height, which is the reason SpaceX gave for moving lower. McDowell expected the Cosmos 1408 fragments to linger for five to ten years, while many of the 3,000 pieces from China's higher 2007 test were still in orbit fourteen years later. Lewis describes crossing the runaway threshold as a gradual loss of safe operations. The permanent runaway lives higher, in the band his model already flags.
The logistics doctrine that survived persistent observation treats each loss as local. On the Ho Chi Minh Trail a destroyed truck stopped moving, and the North Vietnamese shipped enough to arrive net of attrition, as Brief №3 laid out. Proliferated constellations follow the same design, spreading satellites across many planes and replacing losses with new launches. In orbit the destroyed satellite keeps circling in the same shell at orbital speed. Each loss raises the loss rate for everything sharing that shell. Throughput planning net of losses assumes each loss is independent, and a debris cascade makes each loss produce the next.
What This Means
The cascade also serves the attacker after the first impact. Every later collision looks like an accident in a crowded orbit. Investigators then have to separate one deliberate act from hundreds of accidental collisions and close approaches. The author has argued that fragmented attribution is the Russia hypothesis, because an operation succeeds when the case stays open. A debris cascade keeps the case open for as long as the fragments stay in orbit.
The shared picture in orbit already exists. Independent analysts using public orbital data tracked Cosmos 2612's approach for months, the way the hospital's coordinates sat in the database at Kunduz in Brief №1.
The record a hostile swarm calls for has been on paper for years. Moriba Jah's ASTRIAGraph at the University of Texas has combined independent catalogs since 2018 and exposes where they disagree, which is the contradiction as signal of Brief №9. In 2019 MITRE proposed BESTA, a permissioned ledger run by spacefaring nations that compares observed behavior with agreements and stores each anomaly as a tamper-evident evidence docket for adjudication. Guardtime and ESA presented an SST data trust service on a distributed ledger at the 2021 European Conference on Space Debris. The CCSDS Orbit Comprehensive Message, which the US civil traffic system uses in its specifications, already carries planned maneuvers in a standard format.
The rules for a hostile approach exist as proposals for states. At the UN working group on reducing space threats, states raised the need for notification and consultation before rendezvous and proximity operations with foreign satellites, and the US Department of Defense adopted behaviors for its own proximity operations. Those norms address the approaching state. The operator being approached still needs a rule of its own.
Forced maneuvers are a recognized attack. A 2023 AMOS paper on securing space domain awareness lists manipulated tracking data that generates false conjunction warnings and makes an operator spend fuel.
These proposals concern shared data and its integrity. The decision layer sits on top of them. Autonomous avoidance acts on that data at machine speed, and an adversary that shapes the data steers the dodge. Brief №7 answers that shape of problem for software agents: the system that proposes an action commits it through a gate held outside it, under a rule written before the event. Every proposal and every commitment lands on a record a stranger can read. Adoption is the other open problem. BESTA and the ESA service were published as a proposal and a prototype, and the US civil system that would carry shared maneuver plans has faced a proposal to end its funding.
Four questions test whether a constellation is ready for a hostile operator.
When an undeclared object closes inside the screening volume, who decides the response, and under which rule written before the approach?
How long does the conjunction pipeline take to tell a hostile maneuver from a stale orbit estimate?
How many forced avoidance burns can each satellite absorb before propellant reshapes the replacement plan?
Can an outside party verify each maneuver decision from the operator's records alone?