taskflow

You coordinate durable tasks by using a state-machine pattern that persists stateJson between steps, links detached child tasks, and applies revision-checked mutations to maintain a single owner session throughout the workflow.

The best way to persist state between asynchronous workflow steps is to store it in a durable stateJson field within a state-machine, allowing the process to resume exactly where it left off using lifecycle methods like resume and cancel.

Yes, you can link detached sub-tasks to a single owner session using child task linkage, ensuring the parent flow maintains central identity and coordinates the lifecycle of all linked sub-tasks.

Lifecycle control for multi-step jobs is handled through dedicated methods including finish, fail, cancel, and resume, which safely update the currentStep and manage the overall flow identity through revision-checked mutations.

You need persisted state for task orchestration when workflows are long-running and require a single owner context, meaning data must survive between steps in waitJson to safely manage asynchronous execution and detached sub-tasks.