Test the guarantee your dependency advertises
A documented consistency or safety guarantee is a claim, and generating histories and searching them for cycles finds violations of that claim in shipped systems at rates high enough to matter: 1,461 of 13,914 transactions in one measured run, under normal operation.
Do this firstTests bring their own world
Drive the system with concurrent operations, record what clients observed, infer the dependency graph between transactions, and look for cycles that the advertised isolation level forbids.
The reason this is a control and not a research curiosity is where the failures were found. Not under exotic partitions — under normal operation, at about a hundred transactions a second, on a stock configuration. The checker that made this practical handles histories of hundreds of thousands of transactions in tens of seconds, where the previous generation timed out after a few hundred.
The method’s own limit, stated by its authors: it can prove the presence of violations, never their absence. And it reasons about individual objects rather than predicates, so it cannot separate every isolation level from its neighbour.
The decoy
The vendor’s documentation. The system that lost roughly 10% of its transactions to isolation anomalies — with no faults injected at all — was advertised at the time as offering among the strongest data consistency, correctness and safety guarantees of any database available.
Evidence
- Elle: Inferring Isolation Anomalies from Experimental Observations — IIWhere the previous-generation checker timed out or exhausted memory after a few hundred transactions, this approach checked histories of hundreds of thousands in tens of seconds — which is what makes the practice affordable. It revealed anomalies in every system tested.
- Jepsen: MongoDB 4.2.6 — IIAgainst a vendor claim of among the strongest data consistency, correctness and safety guarantees of any database available, the measured result was 1,461 of 13,914 transactions with cyclic dependencies — roughly 10% of transactions exhibiting anomalies during normal operation, without faults injected.
Last reviewed 2026-08-19.