Picture the moment. A hurricane loss lands on a claims adjuster's desk at a regional insurer. The treaty is in force, the premium was paid on time, and the reinsurer's name is right there on the slip. Then someone runs the numbers and the loss stays on the cedant's books, because three structural features of that treaty combined in a way that nobody thought through at placement. No fraud. No bad faith. Just architecture.
This is not a story about fine print.
A wholesale reinsurance treaty is a set of mechanical rules that sort losses into two buckets: yours and ours. The sorting happens automatically, according to structural parameters written into the contract long before any storm makes landfall. The gaps are almost never where cedants expect them, and the cost of finding out late is, in a bad year, existential.
The three levers that leave losses behind
Start with the retention. Every proportional or excess-of-loss treaty specifies a point below which the cedant bears loss alone. On a per-occurrence excess-of-loss structure, that might be $10 million per event. A catastrophe that inflicts $9.8 million in insured losses across the cedant's portfolio touches nothing above that line. The reinsurer collects its premium and pays nothing. For a regional carrier with a concentrated book in a single coastal county, a moderate hurricane making landfall fifty miles from the densest exposure can produce exactly this outcome: a painful, portfolio-level loss that nonetheless falls just below the attachment point. The reinsurer's quarterly result looks fine. The cedant's does not.
The second lever is the occurrence definition, and this is where sophisticated treaty disputes actually live. Most catastrophe treaties attach to a single occurrence, defined by a combination of cause, geography, and a time window, typically 72 or 168 hours depending on the peril. A wildfire that burns for eleven days does not automatically constitute one occurrence under a 72-hour clause. The cedant may be forced to split losses across multiple occurrences, each of which must independently pierce the retention before any recovery is possible. Two separate $8 million loss segments from one continuous fire event, under a $10 million retention, produce zero recovery. The same $16 million aggregate loss, treated as a single occurrence, would have recovered $6 million. That $6 million gap is not a rounding error; for a midsized regional writer it can represent two or three years of underwriting profit on the affected book.
The third lever is sublimits and carve-outs. Many treaties written for cedants with mixed books explicitly exclude or sublimit losses from named perils or specific lines of business. Flood is the most common carve-out in property catastrophe treaties, but others include storm surge, communicable disease, and cyber event aggregation embedded in physical damage claims. A cedant that writes commercial property with broad all-risk wordings may find that the treaty's flood exclusion strips out a substantial portion of a hurricane loss, because storm surge is classified as flood under the reinsurance contract even when the cedant's original policy paid the claim as wind-driven water damage. The cedant took the premium. The cedant pays the claim. The reinsurer, technically, is correct.
Put those three mechanisms together and you get what practitioners sometimes call the gap stack: the retention keeps small-to-moderate events, the occurrence clause fragments borderline events, and the carve-outs remove whole categories of peril from large ones. It is less a safety net than a sieve with three different mesh sizes.
A scenario that makes this concrete
Consider two fictitious but entirely plausible cedants, both buying the same market treaty with a $10 million per-occurrence retention, a 72-hour hurricane clause, and a flood exclusion. Call them Coastal Mutual and Inland Specialty. Coastal Mutual's book is concentrated in a single metro area; Inland Specialty writes a geographically dispersed portfolio across three states.
A slow-moving tropical system drops twelve inches of rain over four days, causing $18 million in combined wind and flood losses for each company. Coastal Mutual's losses are heavily flood-related because its policyholders are near the coast. After the flood carve-out, only $7 million qualifies as a covered peril under the treaty, which falls below the retention. Recovery: zero. Inland Specialty's losses are more wind-driven, with $14 million qualifying as covered, piercing the retention by $4 million. Recovery: $4 million. Same treaty, same storm, same premium. Radically different outcomes, driven entirely by how each portfolio's exposure profile interacts with the treaty's structural parameters.
The lesson here is not that Coastal Mutual was unlucky. It is that Coastal Mutual almost certainly did not model what its treaty actually covered, as opposed to what it assumed the treaty covered. That distinction, invisible at placement, becomes the only thing that matters when the loss is already in the room.
Cedants who have mapped their occurrence definitions against actual exposure concentrations and stress-tested their 72-hour windows against slow-moving storm scenarios tend, in the market's experience, to buy occurrence reinstatements or negotiate manuscript language on flood carve-outs. That costs money. Not doing it costs more.
The common assumption is that buying reinsurance transfers catastrophe risk. The more precise truth is that it transfers the catastrophe risk that survives the treaty's internal sorting machinery, and that machinery was designed, in part, by a counterparty with different interests than yours. Cedants who treat structural parameters as administrative boilerplate are essentially delegating their balance sheet protection to someone else's drafting committee. The sorting will happen either way. The only question is whether you understood the rules before the storm, or after.