a flaw
Every real ceramic, no matter how carefully made, is riddled with tiny imperfections, and it is the worst one of them, not the material as a whole, that decides when the part breaks. Think of the nick at the edge of a sheet of paper: the paper is sound everywhere except that one notch, yet it tears from exactly there. In a ceramic these notches are flaws: pores left between powder particles, voids where an agglomerate failed to sinter, hard foreign inclusions, abnormally large grains, and scratches or chips from machining and handling. A flaw is any crack-like defect that concentrates stress and can start a fracture.
Flaws in engineering ceramics are usually small, from a micron to a few tens of microns, which is often smaller than the eye or even routine inspection can catch. Because a ceramic cannot yield, the flaw that concentrates stress the most is the one that triggers failure, and this single worst defect is called the critical flaw or strength-limiting flaw. It is frequently at the surface, since in bending the surface carries the highest tension, so a machining scratch or a handling chip on the surface is often more dangerous than a larger pore buried safely inside. After a break, fractographers read the fracture surface backwards, following telltale markings to the origin, and there they nearly always find the flaw that started it.
Because strength is set by the single worst flaw, not by an average, the whole game of making strong, reliable ceramics is really the game of controlling flaws: clean, well-dispersed powders to avoid inclusions and agglomerates, careful forming to avoid voids, and gentle finishing to avoid deep scratches. The honest limit is that you can never remove flaws entirely; the goal is to make the worst flaw small and rare. And because the worst flaw varies from part to part in a way you cannot see, two identical-looking pieces can have quite different strengths, which is exactly why ceramic strength scatters and must be described by a statistical distribution rather than a single number.
A batch of silicon-nitride bearing balls all pass a visual check, yet their strengths range from 600 to 1000 MPa. Fractography traces each failure to its origin: the weak ones broke from a 40 micron iron inclusion picked up during milling, the strong ones from mere 10 micron surface pores. Same material, different worst flaw.
Identical-looking parts can differ in strength because each one carries a different worst flaw.
You cannot make a flawless ceramic, only one whose worst flaw is small. And beware: a bigger buried pore is often less dangerous than a smaller sharp surface scratch, because sharpness and surface location concentrate stress more than size alone.