The U.S. Chemical Safety Board has finalized a cluster of major investigations: serious hydrogen fluoride releases at Honeywell’s Geismar, Louisiana facility; the November 2024 explosion at the Givaudan Sense Colour plant in Louisville that killed workers in and around the building; and the October 2024 hydrogen sulfide release at the PEMEX Deer Park refinery that killed two contract workers. Alongside them, the Board’s accidental-release reporting data — now four volumes of incident summaries — shows how frequently serious events occur that never reach a full investigation.

Different chemistry, interchangeable findings

Read together, the final reports repeat a short list of failures that any operating site can audit itself against. Safeguards that existed on paper were degraded, bypassed, or never verified against the scenario that occurred. Process safety information — the drawings, relief calculations, and operating limits decisions depend on — had drifted from the installed plant. Changes to equipment, operations, or staffing were made without a management-of-change review that honestly asked what could now go wrong. And in multiple events, the people at greatest risk were contractors whose exposure the site’s own hazard analysis never fully considered.

The quiet finding: the unit forgot

Beneath the formal causes runs a quieter one: units that had lost the knowledge of their own hazards. The corrosion mechanism was known when the unit was built and unknown to the crew operating it decades later. The reason for an operating limit retired with the engineer who set it. The near-miss that predicted the event lived in a file nobody indexed. Incident investigation keeps rediscovering what organizations once knew — which is an argument for treating knowledge retention as a safeguard in its own right, not an administrative nicety.

CSB reports are free consulting written in other people’s losses. The only cost is taking them seriously before the sequence repeats locally.