Algebraic geometers care deeply about "moduli spaces," which are geometric objects that classify mathematical structures. In this case, the structures being classified are algebraic curves with certain mild singularities, called A_r-stable curves. These are curves allowed to have specific types of singular points, generalizing the classical theory of stable curves. A central goal is to construct a well-behaved geometric space that parametrizes all such curves in a clean, separated (meaning distinct objects correspond to distinct points) way. The challenge is that the natural parameter space, called a moduli stack, is often too large or poorly behaved to directly serve this purpose.
The paper investigates exactly what goes wrong with the full moduli stack of A_r-stable curves. The authors identify two precise technical conditions, called theta-completeness and S-completeness, that a stack must satisfy for a good moduli space to exist. They prove that the full stack fails both conditions, meaning it genuinely cannot support the kind of separated geometric quotient one would want. The obstructions are identified by studying how families of curves behave over certain test spaces, which probe the local geometry of the stack in a very controlled way.
To salvage the situation, the authors carefully carve out a smaller open substack, removing exactly the curves responsible for the bad behavior. They show this substack passes the completeness conditions that the full stack failed, and moreover that it is the largest possible substack with any hope of admitting a separated good moduli space. In other words, nothing larger could work. A companion paper will use these results to prove that this carefully chosen substack does indeed admit a proper good moduli space, at least for small values of the singularity parameter.