The second model ran longer.
That was the good result, and Juno entered it before the failure so the record would not flatten the test into a single discouraging word. The local engineer had changed the distribution between two connected units, using a new coupling to improve energy retention. Under a steady load, the model held beyond the first version’s stable interval. Each contribution remained within its initial watched range.
Then the observer disconnected one.
The remaining unit surged before it could compensate. The output dropped and recovered in a sharp movement that would have been unacceptable in a live barrier. Juno called stop. The local engineer isolated the source, and no one approached until the separate gauge established discharge. They had learned that rule in the previous session and did not need to learn it through an injured hand.
The surface engineer on the channel asked whether the disconnect had been too abrupt.
“It was the contributor’s stop control,” Juno said.
“A real emergency might require a staged withdrawal.”
“Then the model has to provide one, and the contributor has to know the limit before starting.”
