Why is Jenkins still so painful for pipeline as code?

I maintain three CI/CD stacks: Jenkins, GitLab CI, and GitHub Actions. For pipeline-as-code, Jenkins consistently requires 2-3x the configuration lines for equivalent functionality and has higher failure rates due to its declarative/scripted hybrid model.

Key friction points measured in our data engineering context:

* **State Management:** Jenkinsfiles must handle node/agent selection and workspace cleanup explicitly, leading to flaky builds. Our benchmark shows a 12% failure rate for initial pipeline runs on fresh executors vs. <3% for GitLab.
* **Library/Plugin Dependency:** Shared libraries require manual version pinning and global configuration. A plugin update broke our S3 integration for 48 hours.
* **Configuration Verbosity:** A simple dbt build stage comparison.

**Jenkinsfile (Declarative)**
```groovy
pipeline {
agent any
stages {
stage('Build dbt') {
steps {
script {
def venv = 'dbt-env'
sh """
python -m venv ${venv}
. ${venv}/bin/activate
pip install dbt-snowflake
dbt deps
dbt build
"""
}
}
}
}
post { always { cleanWs() } }
}
```

**GitHub Actions**
```yaml
- name: dbt build
run: |
pip install dbt-snowflake
dbt deps
dbt build
```

The modern alternatives provide tighter integration with version control and require less boilerplate for state management. Jenkins' pain points are not about capability, but maintenance overhead and reliability at scale.