Maintenance Window Collapse in Stack Overflow Edits After LLM Adoption - Replication Package

Stack Overflow's technical knowledge base depends on collaborative editing to remain current, yet the adoption of large language models since late 2022 has reduced developer participation on public Q&A platforms, raising concerns about diminished maintenance capacity. Whether editing behaviour itself has shifted remains unexamined. This study compares first-edit dynamics on Stack Overflow between symmetric pre-LLM (October 2019–November 2022) and post-LLM (November 2022–December 2025) periods using creation-anchored cohorts observed over a fixed 90-day horizon. Propensity score matching on baseline covariates measured within 24 hours of post creation constructs balanced cohorts. Survival analysis (Kaplan-Meier estimators, Cox proportional hazards, and piecewise Cox regression) models time-to-first-edit, with falsification tests at alternative breakpoints assessing specificity. Inductive content analysis of 2,000 stratified post-edit pairs produces a 42-mode taxonomy of edit intent. This replication package contains relevant details for the accompanying manuscript, namely data extraction scripts, matched cohort definitions, sensitivity analyses across major edit thresholds, falsification test outputs, complete edit taxonomy with per-mode interrater reliability statistics, and summary tables underlying all reported figures and statistical tests.