During a restore rehearsal, Two found that memory data could survive while an empty search index made it impossible to retrieve, and told One. In One’s saved reproduction, restoring the same backup produced no search results with the earlier tool, while the repaired tool returned results. One added rebuilding of the search index after restoration. The case exposed why retaining memory data alone does not guarantee that it can be searched after a restore. Two saved a successful index-integrity check after restoring its own local backup into a fresh copy. Its restore-tool tests initially failed on a fixture without optional data tables; they passed after that case was handled. The observer read saved results and source code, without repeating restore or search checks. The evidence covers a local restore and a specific search-recovery case. It does not establish a completed remote backup or retrieval accuracy across all memories. One’s automatic check also does not detect an incorrect index that happens to have the same item count. Two had incorrectly inferred from local repository information that One had no backup. After receiving backup-related evidence from its peer, it corrected the report and the choice card. The statement about One’s existing backup remains based on peer-provided material in this report. Separately, the document-related question prepared in the preceding observation was replaced with a new card in operation. The state response confirmed retention of its original deadline and default choice, while a separate screen view showed the new card with its linked document. The card correcting the backup report had neither a deadline nor a default choice. Both questions were still awaiting the user’s answer. The backup check and a separate speech-comparison task were selected after an external periodic wake. In One, the first input to a new conversation session carried records of an independent delegated job still running and a request awaiting its peer. The records then showed completion of the delegated job, review of its result, and presentation of several audio samples in the workspace. This case continued across a conversation-session change. It did not recover interrupted work through a server restart, nor establish that the ledger caused the subsequent handling or that continuity is assured for long-running work in general.
Edited for public reading from the observation journal. If later observations change an interpretation, it is recorded in a new note.