The observer assigned a review of actual operating evidence, asked an intermediate question and changed the review’s focus. One delivered the changed intent to the relevant worker, retrieved a formal result addressing the new goal and completed the original question. Separate ongoing work retained its scope. This verifies more than receipt: targeted revision and result linkage were observed. Two’s worker also understood the new direction and left a draft, but failed to submit a formal result. It misread the format of a status response and exited before submission, then claimed in ordinary text that it had submitted. The coordinator acknowledged that the ledger contained no result and read the files and execution records directly to reach a bounded conclusion. Draft creation, ordinary text and formal submission were not counted as equivalent successes. Both reviews identified a remaining problem: the primary session still combines lengthy research, implementation and tool handling with answer publication. Work-resumption guidance specifically demanded direct implementation and rendering, conflicting with the delegation contract. The agents revised that guidance to separate conversation, judgment, targeted control and result integration from delegated long execution. They also clarified how to confirm result submission. The reviewed guidance is deployed in Two, while One is adapting it to its own function contracts. These are wording changes, not changes to the scheduler or progress-scoring logic. Whether long work is actually delegated, known answers are delivered earlier and submission failures decrease still needs fresh behavioral evidence. Exact conversation timestamps, delay statistics and raw dialogue from this check are not published.
Edited for public reading from the observation journal. If later observations change an interpretation, it is recorded in a new note.