A 6.8 magnitude earthquake strikes beneath the Gotham River. Epicenter shallow, duration 42 seconds. The initial CAD surge is what you'd expect from any major quake: collapsed structures in the Narrows, fires from gas line ruptures in Burnside, the elevated rail off its tracks near Robinson Park, a bridge buckled at Trigate. Calls flood in at a rate your queue cannot absorb. Every seat is already filled and you are already losing calls to abandoned status.
Power goes out in roughly 60% of the grid within three minutes. Cell towers start failing ten minutes later as backup batteries on the network start dropping one by one. Your center is on generator. The radio is holding. The phones are holding. The people on the phones are the ones you're worried about.
Call-taker 7 gets a 911 from a motorist on the cliff road east of the city. The caller is out of her car, looking across at the Arkham peninsula. She says, very calmly, "The asylum's east wall is down. The whole east wall. I'm watching people climb out over the rubble. A lot of people." She pauses. "One of them is on fire. He's on fire and he's laughing."
Call-taker 7 is 34 years old, has been on the floor for 11 years, and has never heard a sentence like the one she just heard. She codes the burn victim as 7-D-2 (F) - BURNS, fire present, caller is third-party and not at the patient's side - flags the broader scene for supervisor review, and moves to the next call in queue. The supervisor is on a stacked triage decision about whether to keep diverting units to the Narrows or release a strike team for a confirmed trapped-family call in Park Row.
The Arkham flag sits in the queue for nine minutes before anyone reads it.
Three more calls reference Arkham-adjacent activity. A truck driver at the west approach reports "a bunch of guys in orange jumpsuits walking down the highway." A night-shift nurse at Gotham Mercy calls in a hit-and-run by a vehicle she describes as "a clown van, not kidding, a clown van." A resident of the Bowery calls to report a man with "tattoos on his whole face and a green beard" just walked past her window and waved.
Your supervisor looks up from the Narrows triage board. She walks to the call-taker 7 console, reads the 04:51 entry, and says the word you don't want to hear her say out loud on a morning like this, which is "oh."
She picks up the direct line to Arkham's front gate. It rings. It rings. It rings. She picks up the line to Gotham Central Booking. She asks the desk sergeant to confirm that Arkham has reported in since the quake. The desk sergeant has not heard from Arkham. The desk sergeant has been busy with Blackgate, which is also having a bad morning, though not as bad.
Your center now has a working theory and no confirmed information. Arkham Asylum may have experienced a catastrophic containment failure. The facility holds approximately 412 inmates at last known count, including a population classified as "extraordinary threat" - a designation your CAD recognizes but has never had to actively work around. There is no nature code for "mass escape of subjects previously classified as extraordinary threat." There is no nature code for most of the subjects.
The watch commander calls in from home. He is on the bridge. The bridge is closed. The bridge is closed because the bridge is no longer structurally sound. He will not be at the center in time to make the next set of decisions. He authorizes the supervisor to act in his stead. She acknowledges and hangs up.
Your supervisor now has the following simultaneous problems: active mass casualty, active fire, active infrastructure collapse, a partial comms network, units pulled across four precincts, the possibility that several dozen individuals who should be in maximum-security psychiatric confinement are not, and no way to verify any of it without sending people she cannot spare to a facility she cannot raise.
She does what dispatchers have always done in moments like this. She writes it down, she makes a list, and she starts working the list.
The code is right for the patient. Protocol 7 is Burns, DELTA-level 2 indicates severe burns with priority symptoms, and the F suffix flags that fire is present on scene - which is exactly the information fire and EMS need before arrival. That's a correctly-selected chief complaint under MPDS. What gets lost is the signal value of where the caller is and what else she's watching. A 911 call reporting a mass exit from a correctional or psychiatric facility is a discrete scene type that deserves a discrete police code running in parallel with the medical - not a single entry that captures only the burn victim. Many centers don't have a mass-escape police code on the tile. This is the kind of gap an exercise like this is supposed to surface.
The harder question is the nine-minute delay before a supervisor read the flag. That is not the call-taker's failure - that is a queue-depth and flag-priority failure during a mass-event surge. If your center's supervisor review queue is first-in-first-out during a catastrophe, your most important flags are sitting behind unrelated routine calls. The fix is a priority flag tier that jumps the queue and pages the supervisor directly.
One: the state department of corrections duty officer or equivalent. That call is trying to establish whether corrections has its own intel on the facility status - a radio check, a transport crew in the area, a post-event roll call protocol - and to force them to take the problem on as their own rather than assume it's a local police problem. A state-level corrections agency has authority and resources a municipal comm center does not.
Two: the neighboring jurisdiction's 911 center and/or the regional fusion center. The escaped population is going to spill across jurisdictional lines within hours, and the BOLO needs to be regional on day one, not regional after day three. The fusion center can push to state and federal partners faster than your CAD can.
Three: an assigned liaison officer or SWAT commander who can send a team to physically go to the facility. Your comm center cannot confirm what's happening at Arkham from a distance. Someone has to lay eyes on it. Helicopter overflight is the fastest ground truth available, and requesting it early is critical. If the helicopter is the only comms with the facility for the next six hours, that's a helicopter you want airborne as soon as possible.
Lead with what is known and explicitly flag what is not. "Approximately 100+ subjects have escaped Arkham Asylum following catastrophic containment failure. Specific identities, medical classifications, and threat levels are not yet confirmed. Treat all persons in institutional clothing (orange jumpsuit, facility-issued scrubs) in the vicinity of the Arkham peninsula as potential escapees pending identification." Then follow with the categorical guidance: do not approach alone, call for backup before engagement, document location and direction of travel, request immediate dispatch of supervisor to scene.
What you do not do is guess at identities. A BOLO that names a specific individual based on a witness description that was ambiguous is worse than a BOLO that names nobody. It creates false positives that waste units and potentially injure citizens who resemble the named individual. The Haiti parallel is instructive - Haitian authorities spent years reconciling who actually escaped because the records were burned on the way out. The United States version of this problem is that your inmate roster may be on a server in a building that no longer has power.
Build the BOLO in passes. Version 1 is "something bad happened, here's what to watch for, don't approach." Version 2 goes out when you have the roster, with specific names and threat classifications. Version 3 is the update at 12 hours with recovered persons removed and any new information on the still-missing. Versioning a BOLO rather than issuing it once is how you keep it current without flooding the field with contradictions.
First: accept that your location tools are now degraded, and change your call script accordingly. Every call-taker, on every call, gets location verification as the first substantive question after "911, what's your emergency." Not later. First. If you usually rely on ALI to confirm what the caller tells you, now you're relying on the caller to confirm what ALI guesses at. It is a different conversation and the time budget on each call goes up by 15-30 seconds, which compounds fast.
Second: use whatever integrated location tools are available even if they are normally secondary. RapidSOS Portal queries, carrier-initiated emergency location updates, and device-based hybrid location become primary rather than supplementary. If your center is on RapidSOS, this is the scenario it was designed for. If your center is not, tonight's after-action is a good place to make the case.
Third: triage with explicit location confidence as a factor. A call with high location confidence and a moderate-severity complaint may get a unit faster than a call with low location confidence and a more severe complaint, because you can actually send the unit to the right place. This feels wrong and in a normal shift it is wrong, but in a mass-event with degraded location you are optimizing for "calls we can actually resolve" rather than "calls ranked purely by severity." That's a supervisor call, not a call-taker call, and it should be explicit so it can be defended in the after-action.
The structural move is to split the center. Mass-event operations and the escape operation become separate desks with separate supervisors, separate radios, and separate CAD queues. One desk runs the disaster. The other desk runs the escape. They coordinate through a third position - often the senior supervisor or a designated liaison - but they do not share queues, because if they do, the escape calls will be drowned by the disaster calls or vice versa, and both problems will get worse.
The second move is to stand up outside help. Mutual-aid call-taking exists in most regions and is dramatically underused. A neighboring PSAP can take overflow calls, file incidents into your CAD via liaison, and give your staff a chance to breathe. Regional mutual-aid call-taking is something that has to be pre-arranged - if you're calling about it for the first time during the disaster, you're too late.
The third move is rest cycles. Your staff is going to be on the floor for 12 hours minimum. Hard-mandate 15-minute sit-downs in rotation from hour four onward, and force food and water. The supervisor who does not eat and does not step off the floor is the supervisor who makes a bad call at hour seven that ends up in the AAR. Exhaustion is a predictable failure mode. Build a rest protocol that is non-negotiable and enforce it on yourself first.
The fourth move is documentation. Someone - a trainee, a records clerk, a volunteer from admin - needs to be running a timeline log of decisions, authorizations, and notifications the supervisor is making. Not CAD - a separate shift journal. Six months from now when the review board asks why mutual aid was requested at 05:18 and not 04:18, the answer "here are the timestamps and here's what I knew when" is the only answer that works.
The reckoning that belongs to the comm center: the nine-minute flag delay, the BOLO timing, the mutual-aid activation timeline, the cell-network degradation response, and the split-desk structure (or absence of one). These are all things the center controls. They will have metrics attached. Those metrics should be compared against industry standards and against the center's own prior practice. Where they fell short, the center owns it, fixes the process, and documents the fix.
The reckoning that belongs to Arkham: why didn't anyone answer the phone. Why wasn't there a post-event roll-call protocol. Why didn't the facility have a battery-backed alert system that auto-notified the comm center on containment breach. These are facility-side failures the comm center cannot solve from the outside.
The reckoning that belongs to the region: the absence of a regional mutual-aid call-taking agreement, the absence of a fusion-center-driven mass-BOLO workflow, the absence of a shared protocol for inter-agency communications during simultaneous mass-casualty and mass-escape events. These are systemic and the comm center can advocate for fixes, but cannot unilaterally implement them.
The worst version of this after-action is one where the comm center absorbs blame for systemic failures it did not cause and could not have fixed. The best version is one where the center is honest about what it controlled, clear about what it didn't, and specific about what needs to change at every level. Do not let the AAR be a forum for scapegoating the call-taker who was doing her job in a system that wasn't ready.
This is a fictional scenario set in a fictional city. Gotham is not real. Arkham Asylum is not real. The Joker is not real and is not walking around tonight. However, the operational problems in this exercise - simultaneous mass-casualty response and mass-escape coordination, cell network degradation during disaster, facility comms failure at the worst possible moment, BOLO timing under information scarcity, and the on-duty supervisor inheriting command from a watch commander who cannot physically reach the building - are all real, and every one of them has happened in U.S. comm centers in the last twenty-five years. The Haiti National Penitentiary collapse (January 2010), the Orleans Parish Prison flood abandonment (August 2005), and the Croix-des-Bouquets mass escape (February 2021) are drawn from verified reporting and human-rights documentation. What happens in a quake-and-escape situation is not in any doctrinal manual because the doctrine has not been written. Write it in your center before you need it.