test(dst): account for re-executed ack-loss retries

[azimafroozeh]

What & why

The ack-loss fleet reports extra edge rows when recovery completes the first insert and the client retries it. The model only counts one insertion. Account for both attempts using the retry result and exact row comparisons, and re-enable dst_ack_loss_client_retry.

Successful retries must remain visible. A retry refused for the original recovery operation cannot justify another insertion. Extra or unrelated rows still fail.

Reconciliation validates exact before/after row candidates and compares their counts explicitly. A detector test covers all zero/one/two-insertion transitions, including recovery losing one of two visible insertions. Row-oracle tests cover both initially absent and existing pairs. Before-reopen capture also covers refused retries; its regression proves the original visible insertion cannot disappear while a second insertion remains forbidden.

Backing issue / RFC

Follow-up to the known harness gap documented in PR 662. No separate accepted backing issue is recorded yet.

Checklist

• Focused change in the DST harness and its existing test owner.
• Regression proved red before the fix; negative oracle tests included.
• Harness report documentation updated. No user-facing behavior change.
• Reviewed against the architectural invariants; no engine or storage protocol change.

Local verification

Commands run from crates/omnigraph-dst:

• cargo test --locked: 30 unit tests, 51 scenarios, lane-B and torn-init passed after the review improvements.
• cargo check --locked -p omnigraph-dst: passed.
• cargo clippy --locked -p omnigraph-dst --all-targets -- -D warnings -W clippy::dbg_macro: passed.
• cargo fmt --all --check: passed.
• September 9 nightly replay: six shards of 60 seeds (226000–226359), 2,810 universes, zero failures. All 79 original failing seeds passed. Ran the built scenarios test binary with dst_fleet --exact --ignored --nocapture, the crate Cargo-config environment, and each shard’s DST_FLEET_SEED_BASE / DST_FLEET_SEEDS=60.

The regression covers seed 79 (final export mismatch) and seed 226251 (one-op arbitration mismatch), each with strict replay and an explicit two-insertion verdict. The full DST suite and all six nightly shards were rerun after the final refused-retry capture fix.

[chatgpt-codex-connector]

Codex Review Summary

This comment shows the latest Codex review activity on this pull request.

Review	Status	Commit	Review trigger
📝 Code Review	✅ Completed 2026-09-09T13:36:34.196098Z	85133a8	PR opened

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review" or "@codex security review".

Codex reacts with 👀 while any review is running, comments if it has suggestions, and reacts with 👍 once all reviews finish with no findings.