the merge that carried nothing

Work was done on parallel branches and collected at a central point. The collection step reported success, and it was not lying: the merge it performed completed without conflict. It also carried none of the work, and nothing in the exit status distinguished that case from a real collection.

The symptom was not an error but an absence. The deliverable expected after a collection simply was not there, and the only thing separating a real collection from an empty one was whether somebody opened the result and looked. It happened often enough to become a pattern rather than an accident, and a phrase began appearing in the record itself to mark the ones that had been repaired after the fact.

The repair each time was to re-apply the branch's content onto the collection point and then verify that the files had landed. Over time the check moved earlier: the collector now confirms that the expected files exist at the destination, rather than trusting that the integration step exited successfully.

It is a small and unglamorous failure, and it wastes more work than most large ones, because the work was finished and simply did not travel. What it proves is that delivery is verified at the destination rather than at the command, and that a successful exit is treated as a claim which needs a second witness.