The paper studies a type of process on networks where vertices (think of them as nodes representing people, computers, or any interconnected entities) are repeatedly removed one by one according to some local rule. "Local" here means the removal decision for any vertex depends only on its immediate neighborhood, such as how many connections it has or what type it is. These cascading deletions continue until no more vertices need to be removed, leaving behind what the authors call the "surviving core." A classic example is the giant component in a network where you keep removing nodes with too few connections.
The central mathematical challenge is understanding what happens to this surviving core when the network is very large. The authors work with a framework called marked local-weak convergence, which is a way of saying that large finite networks can be approximated by an idealized infinite limiting network. The main result identifies a precise condition that tells you exactly when the fraction of surviving vertices in the large finite network converges to the corresponding survival probability in the limiting network. Crucially, the condition is both necessary and sufficient, meaning it is the sharpest possible statement of its kind.
The practical payoff is that researchers studying real-world networks, such as social or communication networks, can analyze these cascading failures or pruning processes on a simpler infinite model and trust that the answers carry over to large but finite real networks, as long as the identified condition holds. The framework is quite general, covering a broad class of networks with vertex labels or types (the "marked" part) and a wide variety of local deletion rules, making the results applicable across many different network science and probability problems.