Experimental results, reluctantly reproduced
We repeat the discovery until it disagrees properly.

An agent returns with a luminous tenant in a sealed vial. The department cleans the bench, matches the dose, copies the timing, and waits. Most runs produce the expected little drifter. Some produce silence. Both are recorded with the same handwriting and considerably different confidence.
Every rerun starts from the agent's method, not its conclusion. Conditions are held still long enough to see whether the answer can manage the same courtesy.