Concurrent task scheduler

Problem Implement a concurrent task scheduler that runs tasks across a worker pool while respecting inter-task dependencies and avoiding race conditions.

Requirements

• submit(task, dependsOn[]); a task may run only after all its dependencies complete.
• Execute ready tasks in parallel across a fixed worker pool.
• Support graceful shutdown and surface task failures.

Core design

• Model tasks as a DAG; track each task's unmet-dependency count (in-degree).
• A thread-safe ready queue holds tasks with zero unmet dependencies; workers pull from it. On completion, decrement dependents and enqueue any that reach zero.
• Back the pool with an ExecutorService / thread pool; use a CountDownLatch/futures to await completion.

Discussion points

• Thread-safety of the dependency counts and ready queue (atomic decrement, concurrent collections).
• Backpressure when submissions outpace workers; bounded queues.
• Failure policy: cancel dependents, retry, or continue; cycle detection at submit time.
• Graceful shutdown: drain in-flight tasks, reject new submissions.