You Cannot Buy Your Way to the Front of the Restoration Queue

By 11:20 a.m. on the fourth day after landfall, the operations director of a 34-person title agency two blocks off Front Beach Road had been awake since five and had read the CISA page on the Telecommunications Service Priority program twice. Her fiber had been dark for 91 hours. The carrier's status page still said restoration in progress, no estimate. TSP described precisely what she wanted: a federal program that directs telecommunications providers to give preferential treatment to enrolled users when their lines need restoring after a disruption. She had a legal pad out and was looking for the enrollment link.

There is no enrollment link for her. There is no version of this where she enrolls, and there was none last April either. That is worth understanding, because the four hours she is about to spend chasing it are four hours she could spend on something that would help.

What TSP actually is, and who runs it

TSP is a real program with a real regulatory basis. It is an FCC program, established under 47 CFR Part 64, Appendix A, and managed by the Cybersecurity and Infrastructure Security Agency at the Department of Homeland Security. It is not marketing and not a carrier product, and it works exactly as described. Vendors restore TSP services in order of restoration priority level — Level 1 first, then Level 2, down to Level 5. A parallel set of provisioning priorities covers new service, running from E for emergency down through 1 to 5.

What it is not is something you can buy, subscribe to, or escalate your way onto during an event. The framework's full name is the Telecommunications Service Priority System for National Security Emergency Preparedness, and that last phrase is not decoration. It is the eligibility test.

The eligibility wall is the entire story

Two facts settle it. The first is sponsorship. The FCC states it plainly: federal sponsorship is required to enroll in the TSP program. You do not apply to your carrier, and you do not apply to CISA as a business. A federal agency has to sponsor the circuit, because the circuit supports a function that agency is responsible for. There is no path from a title agency, a marina, or a 200-seat contact center to that sponsorship, however important the business is to the people who work there.

The second is the category structure. Essential NSEP services sort into four categories, each with a bounded range of priority levels:

• National Security Leadership. Level 1 only. Critical orderwire and control services supporting other NSEP functions. Very few circuits in the country carry it.
• National Security Posture and U.S. Population Attack Warning. Levels 2 through 5. Threat assessment, attack warning, command and control of military forces.
• Public Health, Safety, and Maintenance of Law and Order. Levels 3 through 5. Police, fire, EMS, hospitals. Note the ceiling: a 911 center is generally a Level 3, not a Level 1.
• Public Welfare and Maintenance of National Economic Posture. Levels 4 through 5. Services supporting the national monetary, credit, and financial systems — the systems themselves, not a business that uses them.

The fourth category traps people, because "national economic posture" sounds like it might stretch. It does not stretch to individual commercial customers. And even inside the qualifying categories, the vast majority of enrolled circuits sit at Levels 3 through 5.

The same wall exists on the wireless side. Wireless Priority Service uses a comparable five-level structure with the same NSEP gating, and note what even that program promises: WPS calls do not preempt calls in progress or deny the general public's use of the network. It improves the odds of completing a call through congestion. It does not clear a lane.

None of this is a knock on TSP

The program is well designed for what it does. It solves a genuine problem — making sure that when repair crews have a finite number of hours, the circuits carrying attack warning and emergency dispatch are not sitting behind a retail chain's point-of-sale link in an undifferentiated queue.

There is even a fee structure, which is what makes the misunderstanding so easy. Providers charge a one-time setup fee plus a monthly recurring charge per circuit for restoration priority. So it is not free, and it is still not for sale, because the charge is only available to an authorized NSEP user. Money is not the gate. Mission is.

The instinct that leads a business owner here is reasonable too. Facing a multi-day outage you look for the lever, and there usually is one. This time it exists and is pulled by someone else, for reasons unrelated to what your outage costs you. Even that lever is bounded — CISA's guidance notes that service is not guaranteed on the requested date, only that the provider make a best faith effort. Elsewhere in the same materials sits a line that reads almost as a warning to the person reading it in hour 91: TSP provisioning is not intended to compensate for inadequate planning.

So what does set the order?

Something much less interesting than importance. Restoration follows physical dependency, from the core outward, because there is no other way for it to work.

The numbers from Hurricane Milton make the shape of it visible. On October 12, 2024, the FCC reported 1,012 of 15,878 cell sites in the affected area out of service — 6.4 percent, down from 12.1 percent the day before. Of those, 703 were down for lack of commercial power. Another 221 were down because their transport was cut or dark: the site itself was fine, but the backhaul carrying its traffic to the core was not. Twelve were down because the site was physically damaged. Two days later the count was 341 sites, 2.2 percent, and nearly half were still power.

Read that distribution again, because it is the answer. The overwhelming majority of outages after a landfall are not broken equipment at the customer end or at the tower. They are power and transport. Your circuit is very often intact and entirely useless, waiting on something three layers upstream.

That is why the sequence looks the way it does. Electric utilities restore in a fixed order — hazards cleared, damage assessed, then transmission lines and substations, then neighborhood distribution, then individual service drops. Critical facilities get pulled forward inside that structure. Everyone else is a function of which feeder they sit on.

Fiber restoration runs on the same logic and is gated by the electric one. As one regional fiber operator describes its storm process: restoration always starts from the source of light at the huts and works outward toward the customer, and all power must be cleared before any fiber work is done in an area. Their published timeline puts mainline and feeder repair in days two through five, and individual customer drops at day six and beyond.

So the honest answer to "when does my circuit come back" is: after the substation feeding your node, after the node, after the feeder, after your street is passable, and after a drop crew works down a list you are not on and cannot get on. There is a real priority tier ahead of you — first responders, hospitals, 911 — and carriers say so openly. Below it, the order is topology, not importance.

Your business-class SLA is a billing instrument

This is worth checking rather than believing, because the language is public and consistent across carriers.

A typical business internet SLA commits to something like 99.9 percent availability, a latency ceiling measured between the provider's own points of presence, a packet loss threshold, and restoration within some number of hours of notification. Every one of those carries an exclusion for force majeure — one major carrier's terms list fire, explosion, lightning, power surges or failures, strikes, water, acts of god, and the elements. A hurricane is all of those at once. The SLA is suspended precisely when you most want it.

Even when a credit does apply, look at what it is: capped in that same carrier's terms at one month of the service fee for the affected line. It is a partial refund for service you did not receive. It is not a dispatch priority, and no clause in it obligates anyone to send a truck to you before someone else.

None of this is bad faith on the carrier's part. An SLA is a commercial commitment about steady-state performance, and it was never designed to allocate scarce repair crews during a declared disaster. It is simply not the instrument people think it is.

What actually shortens the outage is procurement, not escalation

The things that work are bought months earlier, and they work for one reason: they put you in a different queue rather than moving you up the same one.

• A physically diverse path, verified rather than assumed. Two circuits from two carriers routinely share a conduit, a building entrance, a bridge crossing, or a splice point. As one network consultancy puts it, two circuits can be sold as diverse and still enter through the same conduit. Verification is concrete: ask for route maps or KMZ files, ask where each circuit physically enters the building, and ask whether the laterals and riser paths are separate. If the carrier will not produce the path, you do not have diversity — you have two invoices.
• A different last-mile medium, not just a second circuit. Two buried fiber runs fail to the same backhoe and the same flooded vault. Fiber plus fixed wireless, or fiber plus cable, fails differently — different plant, different crews, different restoration sequence. The medium is what changes your queue. A second circuit on the same medium mostly changes your bill.
• Cellular failover, sized honestly. LTE and 5G failover is genuinely useful, and it is the most common thing that keeps a small office limping through week one. But the Milton numbers apply to it too — 703 of those downed sites were out for power, and the macro site you fail over to sits on the same grid your building does. Assume degraded throughput on a congested cell, and plan for card readers and email rather than full operations.
• Satellite as a declared stopgap. Ookla's US measurements of Starlink, reported in mid-2025, put the median download around 105 Mbps and median upload near 15 Mbps, with the best state-level median latencies around 38 to 39 milliseconds. Only about 17 percent of tests cleared both the 100 down and 20 up thresholds at once. That is workable for a small office and a poor substitute for a business circuit under load. It also needs sky, power, and somewhere dry to sit — after a storm, three separate problems.

The question worth asking in April

Not "can I get priority," because that answer is settled. The question is: what is the physical path between my building and my carrier's nearest point of presence, and what else shares it?

Most carriers can answer. Most customers never ask — which means the answer usually arrives on day four, from an operations director with a legal pad, at the exact moment it can no longer be acted on.