A second explosion in a week at Motor Oil's Agioi Theodoroi refinery, both during contractor maintenance

On the afternoon of Thursday 15 October 2026, an explosion at Motor Oil’s Agioi Theodoroi refinery near Corinth, Greece — the country’s second-largest — injured four workers, two of them seriously with burns and fractures. According to Greek City Times, the blast followed a gas release during maintenance by an outside contractor crew; the company said the unit involved was out of service for preventive maintenance, and its early assessment pointed to a steam release. No fire was reported. Other coverage placed the work at Unit 750 during a scheduled shutdown, on a high-pressure heat exchanger. The Greek Labour Inspectorate is investigating, and figures for the injured differ between the company’s account and local reports.

It was the second such event at the plant in a week. On 9 October, a separate incident during what Motor Oil described as extensive preventive maintenance killed two contractor workers (Reuters, via The Star). On 17 October, Greek Reporter reported that six people — a contractor and crew members, and Motor Oil personnel — had been detained as prosecutors examined the plant’s safety protocols. Cause and sequence for both events remain under investigation, so what follows stays with what turnaround work makes generically relevant, not with a mechanism that has not been established.

Turnaround as its own hazard state

Both events happened not while the process was running but while it was being worked on. That is the recurring lesson, and it is easy to under-weight: a HAZOP built around normal operating flow examines a node as it runs, while a shutdown puts the same equipment into states the study may treat as footnotes — depressured but not gas-free, isolated by means that can leak, opened for a heat-exchanger bundle pull, with hot work introduced nearby. A reported gas or steam release into an area where welding is underway is exactly the convergence those states produce.

The transferable questions are about scope and controls rather than any one failure. Does the study treat “shutdown,” “line break,” “equipment opened for maintenance,” and “hot work adjacent to hydrocarbon systems” as conditions a node is operated in, with their own causes and safeguards — gas testing and re-testing, positive isolation, confirmation that a line is clean and gas-free before a permit is signed? Are the simultaneous-operations interactions — one crew’s isolation against another crew’s hot work — captured somewhere, given that turnaround work is dense with contractors running in parallel? And where a permit-to-work system is the safeguard credited against ignition, has it been tested against the degraded, time-pressured conditions a turnaround actually runs under? A second event at the same site within days is the kind of precursor pattern that makes those standing questions rather than closed ones.