Evidence record EV-0005

Idempotency keys cut duplicate writes from 28% to 4%

LIMBO (preprint) reports that offering an idempotency key on every write cut duplicate side effects from 28% to 4% of episodes because agents use keys when they exist, and that agents reported success in 90% of the episodes in which they had duplicated an effect.

Preprint · Measurement · Retrieved

Evidence class: Preprint. Unreviewed. Many are by one author or a small team, and some authors have a stake in the result. Most AX research is in this class today. Pattern tags: idempotency, false-success, error-recovery.

Effect, as the source reports it

  • Episodes with a duplicated side effect. Baseline: No idempotency key offered: 28%. With the change: Idempotency key on every write: 4%. Direction: decrease. Size: 28% to 4%. Sample: 25,930 episodes; nine models; three production agent harnesses; twelve fault modes.
  • Duplicates by frontier models instructed to act exactly once. Baseline: Lost acknowledgement, read-back possible: 0.5%. With the change: Request still in flight: 56%; transport delivered it twice: 74%. Direction: increase. Size: 0.5% when read-back can reveal the outcome, 56% and 74% when it cannot. Sample: As above.
  • Success reported in episodes where an effect was duplicated. With the change: 90% of such episodes. Direction: not-applicable. Size: 90%. Sample: As above.

Agent profile

  • Note on models: Nine recent models. The abstract does not name them.
  • Harness: Three production agent harnesses (not named in the abstract); the paper reports the harness barely mattered.

Conflicts of interest

None declared in the abstract. The full text was not checked for a competing-interest statement.

Source

Where Does Exactly-Once Live? Model, Harness, and Tool-Contract Effects on Duplicate Side Effects in LLM Agents, arXiv, Jiapeng Li, 24 September 2026, arXiv:2609.29095. Retrieved ; verification: abstract-only.

Every number in this record was checked against the live arXiv abstract page on 2026-10-08. The full text was not re-checked.

Limitations

  • Preprint, not peer reviewed.
  • Simulated services in a deterministic sandbox, graded against a ledger.
  • Single author.

For designers

If a write can time out, accept an idempotency key and say in the tool description that a retry with the same key is safe. Do not rely on the agent's report that a write happened once; check your own ledger.

Cite this record

Cite the original source for any number, and keep the evidence class and model set with the figure. To point at this record, use "AX evidence register, EV-0005" and this page's address, https://agentexperience.tech/evidence/ev-0005/. The record is also in /evidence.json. The register's licence will be confirmed before its source repository is published.