dispatching-parallel-agents

To debug multiple unrelated test failures concurrently, dispatch separate AI agents to investigate each failing test file or subsystem independently, preserving isolated context to prevent cross-contamination.

Use parallel agents for debugging when you have 2 or more independent failures with unrelated root causes where shared state and cross-interference are unlikely, allowing each agent to work with a clean, purpose-built context.

Prevent context contamination by creating focused, self-contained agent instructions that ensure agents do not inherit unrelated session history, keeping each agent's investigation isolated to its specific failure domain.

Integrate fixes from parallel agents by reviewing each agent's findings, checking for overlapping edits to resolve conflicts, and validating the integration by running a full test suite to confirm all fixes work cleanly.

Parallel agent dispatch is not suited for failures within the same subsystem where shared state is likely. It applies to independent failures across unrelated test files or subsystems where cross-interference is unlikely.

The best way to speed up debugging for multiple independent failing tests is parallel problem-domain dispatch, splitting failures into independent domains and running one agent per domain concurrently instead of sequential debugging.