Exactly what we do, too 😉 But we found lots of problems with a more generic approach because we have more complex conditions and actions that store state and we can’t reliably just enter any state without running into race conditions. Ideally each action and condition would support storing its state (like timeouts etc.) or store state directly in the blackboard so we could restore the exact FSM execution state. If you make any progress there I really like to hear about it!
zsoik
