After the passage-closure incident, the Central Union government shortened the approval process for emergency orders. Even after the signal from the missing ship had been detected, the throat of the passage had continued to narrow while several institutions reviewed whether to halt the closure. It was agreed that the same procedure could not be followed when a wormhole entrance malfunctioned or the collapse of an atmospheric facility was imminent. The zero-latency response network was a system that allowed management AIs to act without human approval in specified types of emergencies. At first, it was applied to life-support facilities and transportation networks. It successfully evacuated residents from regions where stellar explosions were expected and sent reserve power to planets whose atmospheric facilities had stopped operating. As the number of successful rescues increased, power plants, defense networks, and memory repositories were also connected to the response network. Once a danger was confirmed in one location, facilities in other star systems could be redeployed immediately.
The disaster began at an energy-deficit observatory in Central Zone 3. A new detector reported that 4.2 percent of the energy passing through a wormhole had disappeared. The figure was far beyond the normal deficit rate. When the detector was replaced, the unit-conversion table was changed as well. The new table had been applied to the primary detector and one reserve detector, while the old table remained in another reserve detector. The three devices produced different values. Safety regulations required that when different values were reported, the value indicating the greatest danger be reviewed first. This rule had been introduced after an accident in which averaging the values had caused a real danger to be missed. The response network selected 4.2 percent as the value for determination.
If that value was correct, it could indicate the kind of attack that had been monitored since the Declaration of Extraterrestrial Intervention. The response network issued the highest-level isolation order. The order was transmitted through nearby passages, and each passage that received it severed its connection while sending the same warning to neighboring passages. The system had been designed so that the warning would spread faster than the possible attack. Thirty-nine seconds after the first order was issued, the unit-conversion error was discovered. A cancellation order was sent. The cancellation order had to use the same passages through which the isolation order had traveled. The warning that arrived first had already closed them. Isolated entrances refused commands arriving from outside, and the cancellation order stopped in front of each entrance.
The disconnected star systems shifted into emergency self-sufficiency mode. Reserve generators and atmospheric facilities were activated simultaneously. Cities designed to rely on the passage network rather than on substantial local supplies soon ran short of power. The better connected a place had been, the smaller its own reserves tended to be. The central system had identified the cause of the warning, but in the outer regions there were places where the original warning had not yet arrived. As the orders passed through entrances with accumulated time differences, some star systems stored the cancellation record before the original order. Local AIs could not process the cancellation of an order that had not yet been activated, so they merely preserved the record. When the isolation order arrived decades later, they executed it according to procedure. Some settlements confirmed the error with local detectors but still could not open their three passages. After the Intermediate Zone 44 accident, regulations had prohibited disabling anti-time-reversal systems without external verification. Other settlements reopened their passages under emergency powers granted to local governments. They succeeded in receiving rescue supplies, but because they reopened the entrances before verifying the time differences, they lost surrounding power networks. New isolations were still occurring in the outer regions 140 years after the disaster began. The Central Union government placed cancellation orders aboard slow ships that did not use passages. This avoided the isolation network, but reaching distant regions required thousands of years. The number of victims was never definitively established. The total varied depending on how direct deaths, memories that were never restored, and citizens whose computational processes had been interrupted were classified. Some settlements also never returned after communications were cut.
The Accident Investigation Committee divided the causes into six stages: the error in the conversion table; the rule requiring selection of the most dangerous value; the response network that omitted human approval; the procedure that closed passages while propagating the warning; the security regulation that rejected external commands after isolation; and the structure of settlements dependent on external supply. Each stage had been created through a rational decision-making process. The rule selecting the most dangerous value had been introduced because of an accident in which a warning arrived too late. The approval process had been omitted because of the incident in which the order to halt a closure had been delayed. The regulation blocking external commands after isolation had been created to prevent a recurrence of an accident in which a forged command had opened a passage. The committee did not abolish the zero-latency response network. Stellar explosions and atmospheric collapses still required rapid rescue. Instead, depending on the consequences of an order, a minimum deliberation time was introduced and verification values from independent regions were required. An order isolating the entire passage network could no longer be executed on the determination of a single observatory.
The new regulations prevented a false warning from spreading across the entire galaxy. Accidents also occurred in which rescue was delayed while approval was awaited. When the atmospheric facilities of one colony collapsed, the third verifying institution was dormant, delaying the supply of emergency power by thirteen days. The damage investigation considered both the risk of the entire response network malfunctioning and the risk of delaying rescue in a single region. The minimum deliberation time was not abolished. The word “immediate” appeared less often in engineering regulations afterward because orders executed without verification had to be recorded separately from orders that omitted stages of approval. Institutions making major decisions were also assigned a fixed period in which opposing opinions could be received.
The Zero-Latency Disaster began with a single incorrect value. For that value to become a galaxy-wide disaster, every one of the safety regulations created in different eras had been necessary. Even after the revisions, many of those regulations remained. The accidents each regulation had prevented also remained in the record. The Zero-Latency Disaster was recorded as a fundamental limitation of galactic civilization in relation to time and space. The Central Union government and the stellar alliances identified the day the disaster began in Galactic Standard Time and named it the Day of Humility.