MonkeyCode Test Reveals Parallel Patch Risks

A developer tested MonkeyCode's ability to handle parallel coding-agent tasks starting from the same base commit while a human pushes an unrelated commit. The experiment checks whether patches identify their base commit and whether tasks remain isolated. Results show that clear base identification and conflict detection are critical for safe parallel agent use.