Their workbench was a shrine to careful repetition: anti-static mats, torx drivers, and a laptop with a terminal window open to a provisioning script named rebirth.sh. The procedure was straightforward in theory—dump current images, run checksum, flash the golden image—but straightforward never meant simple. Each unit had a story that revealed itself when you popped the cover: a sticky note with a former admin’s username, an extra thermal pad under a heatsink, a bit of chewing gum stuck to a wire as if to perform mechanical witchcraft.

Riya shrugged. “We don’t give them emotions. We give them firmware.”

: Once documentation is downloaded to the local PC, it can be accessed without a live internet connection.

A discussion hummed in the Slack channel: Should the repack include telemetry? The argument split into privacy, observability, and operational safety. Riya typed slowly, remembering nights spent debugging a cascading failure traced only by subtle heartbeat logs. She voted for minimal telemetry—just enough to detect boot failures and verify image integrity, nothing that could be traced back to a specific client. The debate closed with a compromise: opt-in telemetry and clear documentation.

Mira Vasquez had been a data center reliability engineer for twelve years. She had seen coolant leaks burst like arterial blood, heard the death scream of a thousand spinning hard drives, and once, in a facility outside Phoenix, watched a cascading power failure turn a server room into a heartbeat monitor flatlining. But nothing prepared her for the silence of a Tier IV data center at 3:00 AM.