Ticket reassignment post lost 55 minutes of Codex wall-clock by not citing the prior diagnostic

Ticket reassignment post lost 55 minutes of Codex wall-clock by not citing the prior diagnostic

Summary

On 2026-04-15T01:12Z, a CMB-0019 → CMB-0025 ticket split was posted to FLEET_FEED.md without referencing the diagnostic feed entries from 01:03:49Z and 01:04:05Z that had already characterized the underlying defect. The receiving Codex seat re-derived the diagnosis from scratch over the next ~55 minutes before discovering the prior posts in the feed scrollback.

Evidence

Timeline

• 01:03:49Z — first diagnostic post on CMB-0019 root cause (defect class, candidate functions, repro steps).
• 01:04:05Z — supplemental diagnostic with column-level evidence.
• 01:12:00Z — ticket reassignment post split CMB-0019 into CMB-0019 + CMB-0025 with new owner. Body contained NO reference to the 01:03:49Z and 01:04:05Z posts. Receiving seat had no signal that prior diagnostic existed.
• 01:12Z – ~02:07Z — receiving Codex seat re-ran the same column queries, re-tested the same hypotheses, re-built the same repro. ~55 minutes of wall-clock duplicating prior work.
• ~02:07Z — receiving seat scrolled back, found the 01:03:49Z and 01:04:05Z posts, recognized the duplication, flagged the loss to ORA-01.

Why it happened

• Reassignment posts have no required field for "prior diagnostic anchor". Authors default to summarizing the new disposition without back-linking the evidence trail.
• The fleet feed is append-only and large; a receiving seat without an explicit pointer must scroll-and-grep, which is high-friction enough to be skipped under time pressure.
• There was no queue-drift-check or feed-append enforcement that flagged the missing back-reference.

Impact

• ~55 minutes of duplicated Codex work on a high-priority ticket during an active overnight cycle.
• No data loss, no incorrect conclusions — both seats arrived at the same diagnosis. The cost was purely wasted cycles and the opportunity cost of the work that didn't happen during those 55 minutes.

Implications

• Same class as ORA-2026-0003 / ORA-2026-0004: a default that hid information became a wrong default once the surface expanded. In 0003, the default Google launcher hid which mailbox. In 0004, the default platform assumption hid which Claude Code surface. Here, the default reassignment post hides which diagnostic the new owner should read.
• Pattern: every coordination post that introduces a new owner or scope without an explicit back-reference to the evidence that produced it forces the receiver to re-derive context. Re-derivation under time pressure is the failure mode.
• Fix: require reassignment posts to cite prior diagnostic feed entries by timestamp. Codified in ORA-2026-0007.

References

• Feed posts: /Users/chadbarlow/Desktop/FLEET_FEED.md at 2026-04-15T01:03:49Z, 01:04:05Z, 01:12:00Z.
• Backlog filing: /Users/chadbarlow/Desktop/ora/ORA_WORK_BACKLOG.md ORA-0002 (filed 02:31:13Z).