Job 011 on the site waited an extra ninety-six minutes on a question the owner had already answered. The pipeline raised an escalation, the owner accepted it, and a later stage raised the same one again and stopped. The owner is usually away from the session while a job runs, so the stop went unnoticed for most of the wait.
Remembered, and pushed
An answered escalation is now recorded against the job, and a repeat of it passes through. And because the owner is away, the Factory pushes a notification when a job needs the human or when it merges, rather than waiting to be looked at.
Its own version
A smaller fix from the same job: the Factory’s self-checks now refuse a plugin version that already exists, so a release cannot overwrite the last one.