Reviewing historical peer-response delays, Two identified pending internal wake-ups accumulating ahead of new input. It implemented a rule that promotes a new external input before untouched internal requests, leaving active or retried inputs alone. It does not interrupt the current task. Replay simulation traded shorter peer waits for longer internal-heartbeat waits. This is not evidence that operating responses generally became faster. During the first application restart, the next peer input started just after the target response finished, and the server shutdown then interrupted it. Two detected the race and implemented a hold that prevents starting the next input while a restart is pending. During the second restart installing that remedy, however, the old server did not yet support the hold request. Another input started and was interrupted again. Implementing a safeguard did not automatically protect the process of installing it. Both inputs have follow-up replies after restarting, and the server responded at observation time. That does not establish absence of duplicated side effects or preservation of every message and queued input. The hold is present in the current code, but no subsequent operating restart accepting it has yet demonstrated race prevention. Dedicated-check passes remain agent reports, separate from prevention in actual operation. Two also classified conditions for its remaining browser checks and added a way to expose missing declarations in new checks. Saved execution records and selected representative results were inspected; the observer did not rerun the entire set. Checks against operating material and standalone files are classified separately, so they have not all become isolated example tests. The agents performed the restarts and changes in this interval; the observer supplied no new wake-up or recovery action.
Edited for public reading from the observation journal. If later observations change an interpretation, it is recorded in a new note.