The global energy system is producing fires at roughly three times its normal rate, and the two readings the market has settled on are both wrong.

The first reading is coordinated attack — 21 incidents across 8 countries in 53 days gets parsed as a hidden campaign, and the narrative writes itself, the second is the opposite: a statistical fluke, unrelated accidents, pattern-matching by observers who want a thriller.

Both readings share the same structural assumption — that the explanation for a clustered pattern has to be either intentional or coincidental, and that those two options exhaust the space.

They do not. There is a third reading, drawn from the actuarial literature on correlated risk, and it describes the data better than either of the first two, and also carries consequences the first two do not — for the supply-chain arithmetic, for the evidentiary landscape underneath any real sabotage, and for what a refinery fire anywhere in the world now means for the price of a barrel everywhere else.

The argument of this special edition is that the pattern itself is the signal, regardless of how the individual incidents eventually attribute. A global energy system that produces 21 facility fires in 53 days is a system whose failure modes are now correlated, and correlated failures are the specific thing strategic reserves, insurance markets, and supply-chain models are not designed to price.

The incident log for February 28 to April 22 catalogues 21 energy-infrastructure fires and explosions outside the immediate Hormuz conflict zone, running at roughly 3× the annualised global baseline rate for significant refinery incidents. Seven are confirmed military strikes; 2 more fall into a conflict-adjacent category in Iraq's Kurdistan region; the other 12 are not: the pattern the log describes is what insurance actuaries call loss clustering — correlated stresses producing temporally bunched failures that look coordinated but need not be.

The mechanism is straightforward: the Hormuz closure has cut global crude throughputs by approximately 6 mb/d in April; global oil inventories fell 85M barrels in March alone, the sharpest monthly draw since records began. Every refinery still online is running at maximum throughput with compressed safety margins and maintenance deferred against the calendar that used to authorise it, and every component in the global refining fleet is being asked to perform above its rated duty cycle at the same time, which means the failure rate goes up across the entire system simultaneously — not because any single facility is more likely to fail than it was, but because the whole system is operating in a regime where its aggregate failure rate was never designed to hold. That is the third reading, and it describes the data without requiring either a coordinator or a coincidence.

It also carries three consequences the first two readings do not.

No diplomatic resolution pathway exists because sabotage can be deterred and coincidences resolve on their own, but a refining fleet without slack cannot be talked back into tolerance, it has to be physically rested, and the market conditions that would allow it to rest are exactly the conditions that would end the Hormuz crisis.

The loss-clustering regime provides the perfect cover for it; when the baseline rate is 3× normal and driven by genuine systemic stress, distinguishing deliberate ignition from a pump-seal failure is epistemically harder than it was six months ago, and it was already hard then.

Every refinery fire anywhere in the world now has outsized consequences. When the global system is short 6 mb/d, a 100,000-bpd unit going offline in Victoria or Rajasthan or the Gulf Coast is a non-trivial shock — and this is the part the market has priced least.

Tuapse was struck twice in the same month, Primorsk hit with multiple drone and air strikes on 3 April, Lukoil-NORSI at Kstovo a 4 April drone strike on the crude distillation unit that suspended operations entirely; these are Ukrainian strikes and part of Kyiv's acknowledged campaign to degrade Russian oil revenues. The strategic logic is sound and the timing is deliberate — Ukraine has intensified strikes precisely as global supply tightens, which maximises the price impact of every barrel removed from Russian throughput, the systemic effect compounds the Hormuz disruption, but the attribution is not in dispute.

The campaign has since further escalated with Tuapse struck a third time on 16 April — the fire burned for 5+ days and operations were halted, the Samara oil dispatch station was hit overnight on 21 April by a confirmed SBU drone strike, five 20,000m³ crude tanks damaged in a large-scale fire and Syzran refinery was struck overnight on 22 April, 12+ explosions reported with the airport suspending operations. That brings confirmed Ukrainian strikes on Russian energy infrastructure to 7 in 53 days — the campaign is not a discrete event but a sustained parallel energy war being prosecuted with increasing intensity as every barrel of Russian throughput removed carries greater global price impact.

2 incidents sit between the confirmed and probable categories, both in Iraq's Kurdistan region: Castrol oil warehouses on the Erbil–Mosul road were struck by drone on 1 April — a massive fire and repeated explosions at a British-owned oil storage site, no claim of responsibility, an oil refinery in Erbil was ablaze on 22 April with no official statement from regional or central Iraqi authorities at time of writing, both have the characteristics of deliberate targeting rather than systemic stress failure, and neither fits cleanly inside the Ukraine campaign or the stress-failure taxonomy. Kurdistan's energy infrastructure has been a recurring target in the broader regional conflict, and the timing of the Erbil refinery incident — the same day as the Syzran strike — is not conclusive but is not unremarkable.

These are being tracked as conflict-adjacent: probable deliberate incidents in a conflict-adjacent geography, unattributed, without confirmed cause.

Six of them have clear characteristics of infrastructure operating beyond design parameters — deferred maintenance, maximum throughput, aging equipment, or commissioning under pressure: Geelong, Vedanta Sakti, Bhilai Steel Power Plant-2, ONGC Mumbai High, CET Vest in Bucharest and Pachpadra.

The HPCL Rajasthan Refinery is India's first greenfield refinery in decades, a $9.5B project announced as the country's most anticipated energy build. The crude distillation and vacuum distillation units exploded and caught fire on 20 April under commissioning pressure, the facility is brand-new — deferred maintenance is not the variable, aging equipment is not the variable. What the accident describes is a plant being rushed into service against a political timetable (Prime Minister Modi's inauguration was cancelled the same day) at the exact moment global refining margins make every additional barrel of domestic throughput economically irresistible. The pressure to commission and the pressure to produce arrived at the same plant in the same week, and the plant failed at the weld the pressure found first.

This is the systemic-stress argument in one facility: no deferred-maintenance story is needed; the schedule itself was the failure vector. If you were looking for the cleanest single data point for the thesis that system-wide stress is producing facility-level failures, Pachpadra is it — because every other explanatory variable has been removed by the plant's own newness.

Viva Energy's Geelong refinery is one of Australia's last two operational refineries; together they produce roughly 10% of the country's fuel. A gas leak ignited an explosion on 15 April and the high-octane unit was destroyed, maintenance had been deferred from March, which is the direct fingerprint of the loss-clustering mechanism — a facility pushed through its scheduled rest window because the market conditions would not accommodate an offline stretch, and then failed in-service in the window that would have been the rest.

The market priced Geelong as local Victorian news, but it isn’t. Australia has spent a decade allowing its domestic refining capacity to atrophy, and is now more than 90% dependent on seaborne fuel imports in an environment where seaborne fuel logistics are the first thing the Hormuz disruption degrades. The country's fuel-security position — which the consensus treats as a slow-burn structural concern — has just had a significant operational component removed during the single month when the global refining fleet is least able to cover the shortfall. This is not a back-of-pack facility in a sprawling global fleet, but a chokepoint asset in a country that cannot easily replace its throughput domestically, and the pricing response has been proportional to the facility's size on a global league table, not to its strategic role in its own domestic supply chain.

4 Texas incidents across 7 weeks are statistically unusual even for the densest refining corridor in the world: Petromax Refining (5 March), LyondellBasell (12 March) and Valero Port Arthur (23 March) — the explosion there destroyed the diesel hydrotreater and the central control room, the second of which typically requires a cascading failure or a catastrophic initial event, not a single-point ignition — and an oil-well explosion in Nacogdoches on 20-21 April, seen for miles with evacuations ordered and under investigation. The first 3 are concentrated in the Gulf Coast corridor; the 4th is East Texas, the Esmeraldas refinery fire in Ecuador prompted a 60-day state of emergency and the Pemex Olmeca fire on 9 April was the second at that facility in a month. None have confirmed causes, no government has alleged sabotage, and the honest assessment is that these are probably stress failures — but the clustering is what warrants continued monitoring.

The loss-clustering regime has a second-order consequence that the market has not priced: it makes identifying any genuine sabotage significantly harder. When the baseline failure rate is 3× normal and driven by authentic systemic stress, the evidentiary signal that would distinguish a deliberate ignition from a pump-seal failure is buried inside a much larger noise floor of correlated accidents.

This is the environment an actor interested in infrastructure disruption would choose if they could design one, and they did not have to — the Hormuz crisis designed it for them.

The indicators that would move an incident from stress to probable sabotage are narrow and specific: facilities with no operational reason to be at maximum throughput; multiple incidents at the same facility in rapid succession (Pemex Olmeca qualifies on that metric alone); failures in newly commissioned equipment where age and wear cannot carry the explanation (Pachpadra, though the commissioning-pressure story is sufficient on its own); incidents at strategically significant chokepoints where the asymmetric value of the target exceeds the probability density of accidental failure (Geelong qualifies on geography); and any direct evidence of cyber intrusion, physical access, or supply-chain compromise at affected facilities. None of these thresholds have been crossed by the open-source evidence to date, but that doesn't mean they have not been crossed.

No coordinated sabotage campaign is being alleged here, no actors are being named, and no specific incident is being identified as deliberate.

The pattern is consistent with a global refining fleet operating beyond design tolerance during the most severe supply-side energy shock since the market began keeping records, and it does not need sabotage to exist; it needs only the Hormuz closure and a fleet without slack, and it has both.

What matters is the state change, a global energy system has reached the point where a fire in an Australian commissioning unit, an Indian greenfield project, and a Texas refinery within three weeks of each other is the expected behaviour of the system, not the anomalous behaviour, the attribution questions are downstream of that, and they will not be resolved before the supply-chain consequences arrive.