Jepsen: MongoDB 4.2.6
https://jepsen.io/analyses/mongodb-4.2.6
Method
"Using Elle, a new transaction analysis tool developed in collaboration with UC Santa Cruz's Peter Alvaro, Jepsen automatically inferred dependencies between these transactions, and searched for cycles in that graph to identify isolation anomalies."
Population
"clusters of nine Debian 9 nodes"
What it does not show
The report states it did not investigate pauses, crashes or clock skew in depth, nor introduce membership changes or shard reallocation, and describes itself as brief relative to most such reports. The method can prove the presence of bugs but not their absence, and establishes no false-positive or false-negative rate against production incidents.
Kyle Kingsbury (Jepsen)
Against 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.
Tier II: Systematic adversarial evaluation of a shipped system with an automated checker and quantified anomaly counts against a documented guarantee. Observational on one system version, no comparison group.