Deterministic simulation testing is a technique in which every external dependency of a system, such as the network, disks, clocks, and random number generators, is replaced by a simulator, so that running with the same seed always produces the same result. Multiple nodes and failure scenarios are then executed rapidly inside a single process. Any failing run can be reproduced exactly from its seed, which makes rare, nondeterministic concurrency bugs such as race conditions far easier to find and fix.
The distributed database FoundationDB popularized the approach, and systems such as TigerBeetle later adopted it. It is used to inject network partitions, latency changes, and node crashes at random while checking that the system's invariants continue to hold.