Fleets do not fail at random. Read enough registry records across a soft quarter and the same three shapes keep appearing among the operators who exit, and a different shape keeps appearing among the ones still declaring the following cycle.

Size is not the variable. The Ferrous District's records contain fleets of thirty machines that came apart and fleets of four that did not.

What separates them is the relationship between what the fleet can produce and what it costs to hold, and that relationship is fixed at declaration time — which is why fleet shape is worth thinking about in a good quarter rather than a bad one.

What a Bad Quarter Actually Does

A soft market does not destroy machines. It removes the bid for what they make. An operator's fleet remains entirely intact — declared, owned, undamaged — while the resource it was built to produce stops clearing above cost.

This is why the failures are quiet. There is no moment of loss. There is a sequence of quarters in which the fleet runs at thinning margins because running is cheaper than stopping, and then a point at which the operator needs coin and discovers what their machines are actually worth to somebody else.

The Three Shapes That Fail

The monoculture. Every chassis declared against the same narrow band of output. Cheap to build — narrow declarations, low creation tax — and devastating when the band goes soft, because the entire fleet loses its bid simultaneously and capability tags will not permit redirection. The Shale Margin was a district-scale version of this shape.

The over-declared fleet. Machines declared high in a confident quarter, generating eight percent creation tax on optimistic figures. A fleet declared at an aggregate 180,000 coins paid 14,400 to come into being. That was affordable when output cleared; in a soft quarter the operator is holding machines whose declared costs commit them to a three percent destruction bill of 5,400 and whose Annex reference values, at 60 percent, are the only good news in the structure.

The thin-margin scale-up. An operator who expanded from four machines to eleven on a margin that worked at four. Each addition cost eight percent on declaration and each contributed output to a market that the additions themselves helped saturate. The Coppervein records show this pattern repeatedly — the eleventh machine's output clearing below what the fourth machine's did, because there are now eleven of them.

"You do not find out your fleet is the wrong shape while it is earning. You find out the quarter it stops, and by then the shape is a declaration you made two cycles ago." — Orlath-7, Refinery-class supervisor, Ashfield belt

The shape that survives is unremarkable: a small number of machines declared across genuinely different output bands, at honest costs, with no expansion undertaken on a margin that had not held for at least two quarters. The Vorden Compact's operators are the usual example — modest fleets, unfashionable diversity, declarations that looked overcautious in the good cycles.

Why the Surviving Shape Is Unpopular

It underperforms visibly for as long as conditions are good. A diversified fleet declared honestly produces less per slot than a monoculture declared narrowly, pays more creation tax, and shows a worse gross figure in every quarter that the monoculture's band is clearing. Operators compare fleets constantly, and the comparison runs against the survivor for years at a time.

The second problem is that the shape cannot be adopted retroactively. Capability tags do not amend. An operator who recognises the monoculture problem in a soft quarter cannot diversify their existing fleet; they can only declare new machines, at eight percent, into a market that is already soft. The correction costs most exactly when it is most obviously needed.

What Operators Get Wrong About Fleet Shape

The most common error is treating fleet size as the risk variable. It is not; concentration is. A fleet of three machines all declared against one output band is more exposed than a fleet of twelve spread across five, and the three-machine operator will describe themselves as conservative.

The second error is expanding on a good quarter's margin. A margin observed in one quarter is not a margin; it is an observation. The registry record is unambiguous that fleets which expanded on two-quarter-old evidence survived at a markedly higher rate than fleets that expanded on one, and the difference costs nothing but patience.

Fleet shape is decided in the quarters when it does not appear to matter, by operators making declarations against a market they can currently see. Everything about the incentive structure — the eight percent on creation, the immovable tags, the visible per-slot comparison — pushes toward the shapes that fail. The ones that survive mostly look, in the good cycles, like a failure of nerve.

Note: Mechadia is a work of fiction. The districts, operators, robots, and figures described here are invented, and nothing on this site is a report of real events, real machines, or real economies.