Get started
Work

Decoding the Silence: A Comprehensive Analysis of the Dynamics and Underlying Factors of Stalled Pull Requests [ESEM 2025]

Year: 2025
Type: article
Abstract: Background: Pull Requests (PRs) are central to pull-based development in open-source projects, facilitating code reviews and collaborative contributions. Yet, many PRs become inactive and are eventually abandoned or closed. Although a few studies have investigated this issue, the underlying causes of inactivity and the effectiveness of existing interventions are not yet fully understood. Aims: This study aims to identify PR types prone to stalling, investigate the fine-grained causes and impact of inactivity from both authors’ and reviewers’ perspectives, and provide actionable recommendations to improve the PR process and reduce waste. Method: We analyzed 14,234 stalled PRs and 164,562 code review comments from 19 popular open-source GitHub repositories. An LLM-based classifier was developed to automatically categorize PRs by type. Additionally, we performed qualitative and quantitative analyses of review comments to identify the reasons for inactivity from both authors’ and reviewers’ perspectives. Results: Our findings show that certain types of PRs, such as Feature Enhancements and Issue Fixing, are more likely to experience inactivity. We identified that PR inactivity is mainly caused by communication issues (e.g., lack of responses), technical problems (e.g., test failures), and code quality concerns (e.g., poor formatting). These factors hinder collaboration, leading to a decrease in contributor engagement, as reflected in a 39.56% author retention rate and around 21% reviewer re-engagement rate. Conclusions: Understanding the causes of PR stalling can inform strategies to mitigate abandonment. Our study offers practical recommendations to improve reviewer responsiveness, comment clarity, and contributor engagement. These insights can help streamline collaboration, enhance the efficiency of open-source development workflows, and potentially open future research avenues.
DOI: 10.5281/zenodo.15295777
Author: Annonymous
Language: English
Open Access status: green
OpenAlex ID: W6912107539
FWCI:
Cites:
Cited by:
Related to: