On 12 May 2025, a high-end police drone operating as part of a major US city's Drone-as-First-Responder programme crashed on the rooftop of a precinct station in Brooklyn. The aircraft caught fire. The lithium-polymer battery had ejected from the airframe and ignited on the rooftop. A local fire battalion responded with a can extinguisher.
There were no injuries. There was no structural damage beyond the drone itself. By the standards of the drone industry's accumulating fire incidents in 2025 — including the December Jakarta facility fire that killed twenty-two — this is a small story.
What's worth reading is the public statement the operator's CEO put out afterwards.
The forensic detail
The CEO's statement, attributed in the operator's public support documentation under reference SOSB-24-V18, identified the root cause with unusual specificity:
"During high speed flight, the vehicle and battery may experience vibration at the battery connector leading to accumulation of damage with prolonged usage at high speeds. The connector damage will increase resistance between the battery and the vehicle which can lead the battery connector to fail."
In plain English: the contact between the battery and the airframe is a mechanical interface exposed to vibration in the flight envelope the drone is designed to operate in. Vibration accumulates wear at the contact. Wear raises the local electrical resistance. Resistance produces heat. Heat eventually fails the connector.
The same statement classified the rooftop fire as the "first known customer-facing instance" of this failure mode, against a fleet that had logged more than 500,000 flights at the point of disclosure.
Why "first known" is the wrong yardstick
The framing is generous to the customer. For an operator running a five-year fleet plan it's also a number that can quietly mislead.
A fleet logging 500,000 flights against one disclosed rooftop fire reads — to an unhurried planner — as a one-in-five-hundred-thousand event. That reading has two problems.
The first is the failure mode itself. Wear is an accumulator, not a coin-flip. The probability of failure on any given airframe rises with cumulative high-speed-flight time on that specific airframe. The mean fleet age at the point of disclosure is, almost by definition, lower than the mean fleet age in the years that follow. The forward incident rate is structurally higher than the backward one.
The second is what counts as "known". "First known" is bounded by what the operator and customers chose to report. Field engineers across the industry will tell you privately that connector heat-events at the dock are a common-enough source of return-to-base incidents that they get folded into routine maintenance metrics rather than logged as fire-risk near-misses. The denominator on "first known" is the small set of incidents that escalated to a public fire on a public rooftop.
Project the population: a fleet that scales by 10× over the next five years runs five-million flights. Hold the per-flight failure probability constant — generously — and the expected number of rooftop fires is ten. Allow the per-flight probability to rise with cumulative airframe age and the expected count rises further. None of those subsequent fires will be "first known," but each one will land on someone's rooftop, in someone's police precinct, fire station, mining-site supervisor's office, or hospital helipad.
Read the failure-mode statement, not the incident summary. That's where the next five-million flights are.
The connector is the failure surface
In the operator's own words, the connector accumulates damage under prolonged high-speed flight. That's a clean engineering claim, and it generalises beyond this aircraft, this dock, and this operator.
Pogo pins, blade contacts, magnetic dock pads — the whole family of physical-contact charging interfaces in the autonomous-platform stack — share the same sensitivity to vibration, to repeated mate-and-demate cycles, to corrosion, to spring fatigue, and to thermal cycling. The engineering literature documents at least six discrete pogo-pin failure modes: gold-plating corrosion, mechanical wear, spring fatigue, contamination, ESD/over-voltage transients, and thermal-expansion misalignment. OSHA's arc-flash guidance (Pub 4472) explicitly identifies "corrosion of conductive contacts" as a primary risk factor.
Wear, oxidation, contamination, fatigue, ESD — the failures all converge at the connector. Keep the connector in the architecture and you keep the failure surface.
The wireless inductive alternative isn't exotic. Take the contact away and the pogo-pin failure-mode list collapses — there's no mechanical interface for vibration to wear, no metal for humidity to corrode, nothing left for arc-flash regulations to apply to.
What an operator should ask their drone vendor today
Three questions are worth raising in the next vendor review.
- What is your published failure-mode analysis for the contact-charging interface across the dock fleet? Not "is it reliable" but "what fails first, under what conditions, and at what cumulative use." If the answer is hand-wave, the operator is the failure-mode analysis.
- What is your incident-classification policy for connector heat-events that don't escalate to fire? If those events get folded into maintenance, the public failure rate is misleading.
- What does your dock-architecture road-map look like for contactless charging? Most serious vendors are exploring it. The ones that aren't are betting against the regulator.
NOA Wireless's position
NOA designs and manufactures the wireless-power module that sits inside the dock — the picks-and-shovels layer of contactless charging for autonomous platforms. The modules deliver up to 500 W across an air gap of up to 30 mm, at >92% efficiency, with foreign-object detection that cuts power in under two milliseconds. They're sealed against humidity, dust and corrosive environments. They have no exposed metal contacts.
We send a Dev Kit to qualified OEM and integrator engineering teams so the architecture can be tested on-platform before any procurement commitment.
The fire was small. The disclosure language wasn't. Five-million flights are coming, and the connector is still the failure surface.