From Decay to Repair: Reliability Trajectories of Stack Overflow Answers in the Post-LLM Era – Replication Package

Stack Overflow’s community-driven maintenance model faces mounting pressure from declining participation following the mainstream adoption of large language models (LLMs). This study introduces the reliability-trajectory framework to address this gap, sampling 1,500 focal-answer trajectories across four language ecosystems (Python, JavaScript, Java, C#), spanning symmetric pre-LLM and post-LLM observation windows. Each trajectory was manually annotated for failure presence, type, temporal origin, detection channel, repair status, and persistence. Survival analysis, logistic regression, and a 250-case LLM benchmark complemented the annotation. This replication package contains relevant details for the accompanying manuscript. [This is a draft version of the replication package. The final is 10.5281/zenodo.19435257]