FSM threads take random state transitions using PRNG. There's an initial seed 'clusterOptions.seed' sourced from true random device, and everything else is PRNG derived from that initial source (or sub-derivates). This makes test executions in Evergreen reproducible with the initial seed. 

In the context of Antithesis however, this means that Antithesis is unable to influence individual state transitions. Antithesis controls pulls from /dev/urandom, and making it so FSM transitions under Antithesis pull from that source instead would empower Antithesis to explore more effectively.

Antithesis is already a fully reproducible environment, so probably the main issue is how to prevent users from being confused. i.e. trying to run the test locally with the initial seed and expecting the test run to be reproduced.