scheduling

To schedule parallel tasks without lock conflicts, you need a deterministic protocol that checks staleness, enforces risk-level concurrency rules, and acquires path, interface, and resource locks before batch execution. This prevents deadlocks and ensures safe batch construction.

Batch scheduling is a process that selects ready tasks for concurrent execution while enforcing risk-level gating. It applies strict handling and concurrency rules for critical risk tasks, ensuring safe execution without interference between pipeline operations.

To resume incomplete parallel tasks after a session restart, the orchestrator detects incomplete tasks, restores checkpoints, and determines which pipeline steps need re-execution. This recovery handling ensures tasks resume safely without losing progress.

Yes, you can spawn agents for different tasks in the same message during parallel execution. The pipeline interleaving feature allows each task's pipeline to progress independently while agents are spawned concurrently, maintaining safe batch execution.

To prevent deadlocks when running concurrent task batches, enforce path, interface, and resource lock checks before execution. A deterministic protocol ensures safe batch construction by resolving lock conflicts and applying strict handling for critical risk tasks.

You should avoid parallel task scheduling when tasks have unresolved lock conflicts, fail risk-level criteria, or require strict sequential execution. If tasks cannot meet concurrency rules or risk-level gating, parallel batch execution will fail safety guarantees.