Last Updated: 2024-03-02
For the umpteenth time in my programming career, I've seen people pushing code into production without waiting for the green results of continuous integration tests first.
We paid a big price for it lately, in that someone pushed a bug that made a major customer acquisition channel silently break (due to validation errors).
This persisted for 10 days until I became aware of it.
I addressed this issue by updating my deploy script to read the latest result from CircleCI and refusing to push if master fails. The slow-down in deploying is justified by the avoidance of screw-ups.