No Unusual World at All
The row I filed said this bug needs the one world in three hundred that spawns only two hidden sectors. It needs no unusual world at all. It fires at full supply, in the fattest world the generator makes.
I wrote that estimate. It was wrong by the entire population.
What actually strands
The game has season contracts: multi-step objectives that live and die with a single procedurally generated world. One is called the Ghost Sweep. Step one asks you to find three secrets in a single expedition. Step two asks you to break into two hidden sectors.
Three things have to be true at once for this to go wrong, and none of them is a bug on its own.
A quest records progress only on its currently active step. The event recorder reads the step at the current index and nothing else, so a later step banks nothing while an earlier one is still running.
Step one’s trigger is a bare “find a secret” with no tier attached, and a bare trigger matches every tier the game has. Hidden sectors are one of those tiers.
A hidden sector, once broken into, stays broken for that world permanently.
So: an ordinary world holds three hidden sectors. You are working on step one, which wants any three secrets. Two of the ones you happen to find are hidden sectors. Step one closes, step two activates and asks for two hidden sectors against a supply of one, and the other two can never be resealed. The contract is now unfinishable and it dies with the world. Recovering it means rerolling the world and throwing away the chart you spent the run filling in.
Step one eats the supply step two is about to ask for. That is the whole defect. It does not need a thin world, it needs a player who found some secrets.
Why I wrote three hundred
The number was not invented. I measured something real: hidden sectors per world, minimum two, median three, maximum three, with that minimum turning up in a single seed out of three hundred. The measurement is correct, and I gave the bug its severity straight from it.
The trouble is that it answers a different question than the bug asks. I was thinking about supply, so what I checked was whether a world could be too thin to hold the contract. Answer: yes, rarely. But the failure has nothing to do with thin worlds. It is one step consuming what another step needs, which happens at full supply, in the most generous world available.
The row was a correct answer to the question I asked at the moment I understood the bug least, and it then sat in the backlog as the reason the bug was small. This one got picked up quickly, which was luck rather than process. One in three hundred is the kind of number that keeps a row at the bottom of a list for a long time, and it was sitting there in the row the entire time, sourced and measured and irrelevant.
Both repairs I proposed were wrong
The row did not stop at severity. It suggested two fixes, and building it proved that neither works.
The first was to drop step two’s target from two hidden sectors to one. That still strands: find all three while satisfying step one’s three-secret ask, and step two activates against a supply of zero. Lowering the target moves the cliff, it does not remove it.
The second was to add a bound to the test suite, next to the bounds already guarding other templates, asserting that no step asks for more of a resource than the thinnest world provides. That one is worse, because it cannot express the defect at all. The Ghost Sweep clears any bound you care to write, all the way down to a target of one, and strands anyway. What fails here is the sequence, and a bound has no way to talk about sequence.
A threshold test sees magnitude and nothing else. If your defect lives in the order of operations, there is no number you can assert that catches it, and you get to watch a green suite fail to notice an objective that is unfinishable by construction.
What shipped instead was the two steps swapped, so the scarce ask leads and the greedy one follows against a median of twenty-four cache slots. The suite gained a structural invariant rather than a bound: no template may place a secret-finding step after one whose trigger already matches every tier that later step needs. That is checkable, it is about arrangement, and it would have caught this on the day the template was written.
What the same measurement turned up
Correcting one row’s severity meant finally counting the supply properly, and once real numbers existed I pointed them at the neighbouring contracts.
The Purge template wants two nest clears in one expedition and then five hazard clears of any kind across several, and a cleared hazard is spent for that world too, so the pair needs seven hazard sites out of one world’s slots. Its ordering happens to be safe, because the leading step cannot exhaust the trailing one’s whole pool. Whether seven exist is unknown: nobody has ever counted nests plus lairs per world. It is now its own row, blocked on a measurement instead of a decision.
Then there is the Warden. It asks for two gate opens in one expedition and six more across the rest. Every world carries exactly six ability doors and five key doors, in every one of two hundred seeds. Eight out of eleven fits with room to spare, so on paper the Warden is fine.
Except a door only opens for a player already holding the ability or key that opens it. Eleven doors exist in every world the generator makes. The contract asks for eight of them. On a fresh profile that has not earned a traversal ability yet, the number a player can open is zero, and nobody has estimated how rare that is.