board: coresearch map — refuse, not renumber, for a single-owner Lance log

[AdaWorldAPI]

What this is

One board entry: the /coresearch exploration map for HubSPO-rs ISS-LOG-RACING-WRITER. That issue is the racing-writer gap a codex review found on #1466. The map adopts nothing; the operator chooses what goes forward.

• Entry: .claude/board/entries/2026-10-10-coresearch-single-writer-lance-log.md
• Index: regenerated with entries_index.py --write, plus one curated finding cell.

What the map says

• Lance 13 renumbers every racing Append. This was read in the 13.0.0 source:
  • Append against Append is compatible.
  • Strict mode is Overwrite-only.
  • Every attempt rebases first.
  • D-LNC-5b measured the same on 11 and 12.
• Delta and Iceberg behave the same way. Refuse-not-renumber belongs to the retry policy over a refusing primitive.
• A public extension point could refuse instead, without patching Lance. It is the WriteParams.commit_handler wrapper that pins the expected slot. It is CLAIMED until probe P-SLOT-1 runs, and it carries named conditions: delegate every trait method, refuse with OtherError, build the inner handler from an allowlist, and treat it as temporary until lance#8699.
• Hazards found in today's HubSPO-rs log:
  • It mis-names a race as Corrupt.
  • Diverged means both "nothing written" and "durable".
  • Any non-batch Lance version wedges it.
  • "Never compacted" does not cover cleanup_old_versions.
• Corrections to prior art, recorded in the entry; the owning files are not edited here:
  • The F6 "verified identity" holds only under one writer.
  • find_frame returns the first row, unordered.
  • D-LNC-3's fence-falsifier expectation is false for a plain Append on 13.
  • The MemWAL writer_epoch claim describes WAL fencing, not dataset-commit fencing.

Checks

• entries_index.py --check passes.
• citation_decay.py --since origin/main reports 0 new decays.
• No protected append-only file is touched.
• No code changes.

🤖 Generated with Claude Code

https://claude.ai/code/session_01PPcakjk7ioQsnasmNDtc5z

---

Generated by Claude Code

[coderabbitai]

Important

Draft PR not reviewed

Draft PRs are not automatically reviewed by default.

• Trigger a manual review

To automatically review draft PRs, update your CodeRabbit configuration:

reviews:
  auto_review:
    drafts: true

• Autofix · Keep fixing CodeRabbit findings and required CI, and resolving merge conflicts

Warning

Billing warning: we have not been able to collect payment for this subscription for more than 72 hours. Please update the payment method or pay any pending invoices in Billing to avoid service interruption.

---

_{Comment @coderabbitai help to get the list of available commands.}

[AdaWorldAPI]

Closed unmerged. The case this map explores, two writers appending to one mailbox's log, is excluded by SoA ownership: one writer per mailbox. It is a broken deployment, not something the log handles. It was never observed in a run. HubSPO-rs records it as excluded, and the one single-writer finding moves to its own entry there.

---

Generated by Claude Code