Bureaucracy Times

All the news fit to be declassified.

Category: Agency Policy & Oversight

  • Eleven Safeguards, One Wrong Valve, and a $95 Million Explosion at Shell

    A worker clicked the wrong valve.

    Six minutes later, a furnace exploded.

    That is the shortest possible version of what happened at Shell Polymers Monaca in Pennsylvania on June 4, 2025. But a newly released investigation by the U.S. Chemical Safety and Hazard Investigation Board makes clear that reducing the accident to human error would miss almost everything important about it.

    The CSB found that Shell had identified the possibility of a potentially fatal explosion, had 11 administrative controls that it believed would prevent such an event, and had access to engineered protections that could have made the mistake substantially harder to commit.

    The 11 controls failed. Furnace 5 exploded. Shell estimated the property damage at $95 million.

    Remarkably, nobody was injured.

    The Setup: A First-Time Task, a Confusing Screen

    The Monaca complex, which began operating in 2022, converts ethane into ethylene for plastics production. Its ethane cracking unit contains seven furnaces.

    Furnace 5 had been taken offline so workers could clean its coke trap. None of the facility’s seven coke traps had previously been cleaned since operations began. When the work was finished, Shell began returning Furnace 5 to operation.

    That required reopening large motor-operated valves separating the furnace from the downstream process.

    The engineer assigned to open one of those valves had never performed the task before.

    The computer interface did not make matters easier.

    Three valves appeared on the same safety-system logic screen. Their identifiers differed only by their final digit: 511, 512 and 513. The furnace-side valve was displayed near the top of the screen; the tower-side valve the engineer intended to open was near the bottom.

    After the engineer entered the system’s debug mode, the screen refreshed and returned to the top.

    The engineer inadvertently commanded the wrong valve open.

    The Alarm That Told Nobody Anything

    Even then, the system had another opportunity to reveal the mistake. An alarm activated when the unexpected valve began moving. But its description was essentially identical to that of the expected valve, again differing principally by the last digit. Because operators knew a valve was supposed to be opening, the alarm was acknowledged without corrective action.

    Eventually the intended valve was opened as well.

    That created a path allowing flammable cracked gas to flow backward into Furnace 5, where the pilots were still burning.

    At 2:21 p.m., the gas ignited.

    The explosion severely damaged the furnace and released an estimated 5,100 pounds of ethylene and combustion products. Fifteen employees were evacuated. A contractor became trapped in an elevator beside Furnace 5 and had to be rescued. Other workers were also in the vicinity.

    There was no exclusion zone around the furnace because Shell did not consider the operation a startup activity or abnormal situation.

    They Knew. They Chose Paperwork Anyway.

    The more consequential finding, however, concerns what Shell knew before the explosion.

    According to the CSB, Shell’s process hazard analysis had already identified cracked-gas backflow as a potentially fatal explosion scenario.

    Shell nevertheless relied on 11 administrative controls — things dependent upon policies, procedures, alarms, training and workers correctly carrying them out — rather than an engineered control capable of physically or automatically preventing the dangerous valve configuration.

    The report becomes particularly uncomfortable when it examines Shell’s earlier risk analysis.

    Shell had identified two administrative safeguards against an explosion: operator response to a high-methane alarm and a startup procedure.

    Neither was functioning as assumed on the day of the accident.

    The alarm was suppressed, and the startup procedure considered during the hazard analysis wasn’t being used.

    The Protections Were Right There

    Shell’s hazard-analysis team had previously concluded that adding engineering controls would be “grossly disproportional to the risk reduction” they would provide. CSB reached a rather different conclusion, finding that Shell had actively chosen to rely solely upon administrative controls to prevent a potentially fatal explosion.

    Perhaps most strikingly, engineered protections were not some technology waiting to be invented.

    Shell licensed the furnace technology from Linde, whose original design incorporated engineered safeguards. One involved a sequenced-key interlock intended to prevent an incorrect valve opening. Another was designed to automatically close the furnace-side valve when conditions associated with backflow were detected.

    Those protections were not implemented for the double-isolated configuration involved in the accident.

    Nineteen Times It Worked. Once It Didn’t.

    Meanwhile, Shell had successfully performed essentially the same isolation-and-return operation 19 times before the explosion.

    Nineteen successes apparently demonstrated that the procedure could work.

    The twentieth demonstrated what happened when it didn’t.

    $26,480 in Fines. $95 Million in Damage.

    Afterward, Shell changed the system. Operations personnel can now manipulate the valves from a local control panel rather than requiring a process-control engineer to override the safety system. New logic prevents the tower-side valve from opening if the furnace-side valve has mistakenly been opened, and Shell developed a new step-by-step operating procedure.

    OSHA also issued three citations stemming from the accident. They addressed shortcomings in the process hazard analysis, inadequate consideration of human factors and the absence of clear operating procedures for returning the valves to normal operation.

    The proposed penalty was $26,480.

    The estimated property damage was $95 million.

    The CSB’s report therefore provides a useful distinction between the person who made the final mistake and the system that determined what would happen when someone inevitably made one.

    A worker selected the wrong valve.

    But before that click ever occurred, the explosion hazard had been identified, administrative safeguards had been chosen, engineered protections had gone unused in this configuration, alarms could be suppressed, an inexperienced engineer could manipulate nearly identical controls, and the interface provided precious little assistance in distinguishing one valve from another.

    Human error explains the click.

    It does not explain the system that allowed one wrong click to become a $95 million explosion.

  • The Government Breached Its Nuclear Promise. A Federal Court Just Said It Can’t Bill Ratepayers for the Delay.

    For more than 40 years, the Department of Energy has failed to collect spent nuclear fuel it legally agreed to remove. The latest court ruling closes off its most creative escape route yet.

    In 1982, Congress struck a deal with the nuclear power industry: utilities would pay into a federal waste fund, and the Department of Energy would permanently dispose of their spent nuclear fuel. The deadline was January 31, 1998.

    That deadline passed 28 years ago. DOE has not removed a single rod of spent fuel from any power plant. There is still no approved permanent repository. Yucca Mountain, the designated site, remains mired in a political stalemate with no resolution in sight.

    On September 4, 2026, the U.S. Court of Appeals for the Federal Circuit handed down its fifth ruling in this long-running series of cases, affirming a $145 million judgment against the United States in favor of three decommissioned New England nuclear utilities: Connecticut Yankee Atomic Power Company, Maine Yankee Atomic Power Company, and Yankee Atomic Electric Company, collectively known as “the Yankees.”

    The government did not dispute that it had breached its contract. It argued instead that it shouldn’t have to pay.

    The Arithmetic of an Ongoing Breach

    The Yankees stopped generating electricity in 1996 and completed decommissioning their plants by 2007. They continue to exist for one reason: the federal government has not collected the spent nuclear fuel sitting at their sites, stored in steel casks on concrete pads known as Independent Spent Fuel Storage Installations, or ISFSIs.

    Maintaining those ISFSIs costs money. Security, radiation monitoring, regulatory compliance, licensing. During the 2017 to 2021 claim period at the center of this appeal, those costs totaled $145 million. Because DOE has insisted it still intends to eventually perform under the contract, courts have treated the breach as “partial” and ongoing. That status prevents the Yankees from seeking a final, comprehensive damages award; instead, the litigation proceeds in five-year installments. This appeal was the fifth round.

    Four prior rounds produced nearly $500 million in judgments. Approximately $396 million of that has been returned to ratepayers.

    The Government’s Argument: Your Investment Gains Are Our Defense

    Federal regulations require nuclear utilities to maintain Nuclear Decommissioning Trusts (NDTs), segregated funds built from ratepayer contributions, to cover decommissioning and ongoing SNF storage costs. The Yankees’ NDTs have received $405 million in ratepayer contributions since their establishment in the early 1980s and have generated $339 million in investment earnings since 2009 alone.

    During the 2017 to 2021 claim period, the NDTs earned nearly $185 million in investment gains, more than enough to cover the $145 million in ISFSI expenses.

    The government’s argument was straightforward, if audacious: the Yankees used NDT funds to pay their breach-related costs, and those funds earned $185 million in that same period. Award them $145 million in damages on top of that, the government said, and you’re giving them a windfall. The NDT investment gains should offset the damages to zero.

    The Court of Federal Claims rejected that argument in February 2024. The Federal Circuit, in an opinion authored by Judge Stark and joined by Judges Lourie and Prost, agreed.

    Why the Argument Failed

    The Federal Circuit identified two independent reasons the government’s offset theory doesn’t hold up.

    The gains didn’t mitigate anything. Mitigation, in contract law, means taking actions that actually reduce or avoid a loss. The Yankees’ ISFSI expenses did not go down because the NDTs had a good investment year. The costs mounted regardless. The government couldn’t identify a single expense that decreased as a result of the NDT gains. A funding source that pays for costs without reducing them is a checkbook, not mitigation.

    The court drew a sharp contrast with prior SNF cases where offsets were appropriate: utilities that upgraded to higher-enrichment fuel assemblies and saved money per cycle; utilities that reduced wet storage fees; utilities that pursued off-site storage arrangements. In each of those cases, the non-breaching party took action that actually reduced the damage caused by the breach. The Yankees’ NDT investment returns did neither.

    The gains aren’t connected to the breach. The mitigation offset rule requires a direct relationship, in time and subject matter, between the breach and the event the government wants to use as an offset. The NDTs were established in 1984, more than a decade before DOE’s breach began in 1998. They were created because federal law required them, not as a response to anything DOE did or failed to do. The court rejected the government’s argument that the relevant moment is the claim period rather than when the trusts were created, finding no legal authority supporting that reframing.

    The investment gains, the court wrote, are a “remote consequence” of the breach. A favorable one, but not one that reduces the government’s liability.

    The Windfall That Isn’t

    The government’s windfall argument had intuitive appeal: if the Yankees have $185 million sitting in their trust accounts and only owe $145 million in expenses, why should taxpayers also cut them a $145 million check?

    The court’s answer was precise: because the money in those trusts doesn’t belong to the Yankees. The NDTs are held in trust for ratepayers. Federal regulation requires that when DOE finally performs under the Standard Contract and the Yankees’ SNF storage obligations end, the NDT funds, including investment gains, must be returned to the ratepayers who contributed them. When the Yankees draw on NDTs to cover breach-related costs, they are borrowing against a fund that belongs to someone else. The damages award reimburses that fund.

    Reducing the damages award to zero, as the government asked, would leave ratepayers $145 million short. That is not preventing a windfall. That is engineering one, for the government.

    The Symmetry Problem

    The court added a final reason rooted in logical consistency.

    If NDT investment gains offset damages, NDT investment losses must increase them. That is not a novel legal principle; it is arithmetic. The same activity cannot be mitigation only when it is profitable. At oral argument, the government strongly implied it would never accept liability for NDT losses during partial breach periods. The court held that position disqualifies the government from claiming credit for the gains.

    “The NDTs and their investments are either mitigation activities or they are not,” the court wrote, citing prior Federal Circuit precedent. The government chose to disclaim responsibility for the downside. It cannot benefit from the upside.

    Where Things Stand

    The $145 million judgment stands. The Yankees, which exist solely to store nuclear waste the federal government contracted to remove a quarter-century ago, will receive that amount, which flows back to the ratepayers who funded the trust in the first place.

    DOE has no timeline for SNF removal. Yucca Mountain has no funding, no NRC license, and no political pathway. The courts have now awarded nearly $650 million in damages across five rounds of litigation, with more claim periods accumulating and more rounds ahead.

    The Standard Contract was supposed to solve the national problem of permanent nuclear waste disposal. Instead, it has become a perpetual litigation machine. The government loses, pays, and then tries to find new ways not to pay in the next round.

    This was the fifth round. It will not be the last.

    Connecticut Yankee Atomic Power Company, et al. v. United States, No. 25-1395 (Fed. Cir. Sept. 4, 2026). Panel: Lourie, Prost, and Stark, Circuit Judges. Opinion by Judge Stark.

    Fediverse Reactions

  • 128 Out of 133 VA Medical Centers Can’t Schedule a Mental Health Appointment Within Seven Days

    Seven Days, Five Systems, and a Community Care Promise Still on Paper

    The Department of Veterans Affairs has a seven day standard for one very specific step in mental health care. When a VA clinician refers a veteran to the community care program for mental health services, the consult is supposed to be scheduled within seven days. In fiscal year 2025, according to the VA Office of Inspector General, five of 133 VA healthcare systems met that standard. The other 128 did not.

    The community care program exists because the VA concluded, years ago, that it could not meet all veterans health care needs inside its own walls. Congress passed the VA MISSION Act to formalize that decision and to expand veterans access to care through outside providers.

    The model is straightforward: when VA cannot provide timely or appropriate care, it buys that care from the community through contracted administrators. The premise is sound. The execution, as laid out in VA OIG Report 25 03623 147 covering October 2024 through September 2025, is lagging behind the VA standard the department set for itself.

    How community care mental health scheduling is supposed to work

    The referral chain on paper is not that complicated. A VA provider decides a veteran needs mental health care that is better delivered outside the VA system. That may be because of local capacity, specialty needs, or geography. The provider enters a community care consult in the VA system. From there, the case routes to one of two third party administrators that manage the VA Community Care Network. Veterans in Regions 4 and 5 go to TriWest Healthcare Alliance, while veterans in Regions 1 through 3 go to Optum Public Sector Solutions.

    The contractor is responsible for taking that consult and turning it into an appointment with a community mental health provider. The job sounds mundane: 1) reach the veteran, 2) confirm preferences, 3) find an in network provider with capacity, and 4) schedule a date and time.

    The VA tracks two separate metrics for this process. The first is the seven day standard for how quickly the consult is scheduled. The second is a 30 day contract metric for when the appointment itself is supposed to occur.

    Those are different thresholds that measure different parts of the pipeline, but both matter. If the consult stalls before it is scheduled, the clock on the 30 day metric never really starts from the veteran point of view.

    The OIG review found that the system frequently slowed down or stopped at the scheduling step, where staff tried to reach veterans, confirm provider availability, and reconcile preferences with real world options.

    What the OIG actually found

    The headline number in the August 10, 2026 report is blunt. One hundred twenty-eight of 133 VA healthcare systems averaged more than seven days to schedule community mental health consults in fiscal 2025. Five systems met the standard. That is not a near miss. It is a system wide miss on a metric the VA defined as its own expectation for timely scheduling.

    The picture is more mixed on the 30 day contract metric for when appointments occurred. The OIG found that approximately 80 percent of community mental health appointments happened within 30 days. For the remaining 20 percent or so that did not meet that benchmark, veterans waited an average of 56 days for the actual appointment. That 56 day figure does not describe all community mental health visits. It describes the subset where the system already failed the 30 day standard.

    The OIG did not attribute that gap to a single dramatic breakdown. It pointed instead to a cluster of mundane but consequential problems. First, VA staff and contractors relied heavily on phone calls and postal mail to reach veterans. In many cases, they did not use more efficient or reliable digital contact methods that could shorten back and forth.

    Second, community providers did not always offer appointment types that matched what veterans wanted or needed. Preferences about in person visits versus telehealth, or about certain modalities, were not consistently captured in systems in a way that schedulers could act on. That mismatch produced more calls, more rescheduling, and more delay.

    Third, schedulers reported difficulty getting timely responses from community providers to confirm availability and lock in appointment slots. Unreturned calls, limited office hours, and fragmented communication channels all translated into longer waits between referral and scheduled date.

    A fourth problem sat on top of that. Some community providers insisted on duplicate documentation that the VA had already included in the original referral package. That meant more requests back to VA facilities for forms and records the system had already sent once.

    None of this would impress a management consultant. There is no exotic algorithmic failure here. The VA set a seven day standard for scheduling. The system repeatedly tripped over phone tag, partial preference data, hard to reach providers, and redundant paperwork.

    The structural problem behind the scheduling problem

    The OIG findings land in a landscape that was already crowded with warnings about the VA community mental health network. In November 2024, the Government Accountability Office reported in GAO 24 106410 that the VA assessment of its community care network adequacy for mental health needed improvement. GAO highlighted gaps in how VA evaluated whether the network had enough providers to meet demand.

    An April 2024 VA OIG audit on network adequacy and contractor performance came to a similar conclusion. It found that oversight of the community care network and of contractor performance needed strengthening, including how VA evaluated whether contractors met access standards. On paper, VA relies on TriWest and Optum to maintain a network that can actually deliver the care veterans are referred to receive.

    The provider directories that veterans and schedulers depend on have their own problems. A 2022 GAO report, GAO 23 105290, found that veterans sometimes could not book appointments with providers who appeared in the Community Care Network directory as available. Inaccurate entries, providers who were not accepting new patients, and outdated information all contributed.

    Viewed against that backdrop, the FY2025 scheduling failures are not a stand alone surprise. They are downstream of a network with thin provider supply in some regions, directories that do not always match reality, and oversight that has not kept pace with the size and importance of the community care program. You can streamline the scheduling workflow, and the OIG suggests ways to do that. You still cannot schedule an appointment with a provider who is not actually available.

    The six recommendations and the scorecard so far

    The August 2026 OIG report made six recommendations to the Under Secretary for Health. Two have already been closed as implemented. One directed VA to consider contractual adjustments with third party administrators to improve timeliness of community mental health scheduling. The other required the department to issue standardized guidance that makes clear community providers should not request documentation that VA already includes in referral packets.

    Four recommendations remain open. The first calls for VA to identify and share best practices for contacting veterans in ways that actually reach them quickly. The second focuses on ensuring scheduling systems capture and use the full range of veteran preferences when matching to appointments. The third directs VA to evaluate whether TPA network adequacy standards are sufficient for both in person and telehealth mental health providers, given demand. The fourth urges a review of ways to increase community provider use of VA systems so that communication and documentation flow through consistent channels instead of scattered faxes and phone calls.

    The two closed recommendations nibble at the edges of process friction. The four that remain open go to the structure of how VA, its contractors, and its provider networks actually operate. Whether those structural fixes move from recommendation to enforcement before the next oversight report arrives is an open question.

    Why the delay metrics matter

    Mental health care for veterans who reach the point of a community referral is not elective. A VA clinician has already determined that care is needed and that the department cannot provide it internally in the required way or time frame. For the roughly one in five community mental health appointments that missed the 30 day contract metric in FY2025, the average 56 day wait was not a trivial inconvenience. It represented weeks in which needed care did not occur.

    The veteran population relying on these services carries a documented and elevated risk of suicide and other serious mental health outcomes. That is part of the reason the VA adopted a seven day scheduling standard in the first place. It reflects an institutional understanding that long gaps between referral and first appointment are dangerous.

    The VA, GAO, and the OIG have been writing versions of this story for years. Reports on scheduling, network adequacy, directory accuracy, and contractor oversight repeat the same themes with new data and new report numbers.

    The 25 03623 147 review does not reveal a hidden crisis. It documents, again, a pattern of missed standards, modest corrective steps, and partial follow through. Meanwhile, the metric on the books remains seven days. Five out of 133 systems met it.

    The scheduling crisis documented here is one thread in a larger pattern of VA care access failures. The $1.7 billion blind spot in VHA eligibility determinations shows how the same system sends hundreds of thousands of veterans to outside providers they’re ineligible for, while simultaneously failing to inform eligible veterans they have any outside option at all. The Mountain Home test result crisis illustrates what happens inside the facilities where veterans land when those referrals don’t materialize. And for veterans whose caregivers depend on the VA’s program stability, the 2028 caregiver sunset is the next cliff approaching.