OrchestraBench is a benchmark for evaluating the failure modes, recovery, and decomposition quality of multi-agent orchestration frameworks. It assesses pipeline failures and recoveries through a controlled failure-injection harness and introduces cascade radius and per-failure-mode recovery as primary metrics. On a 26-case gold-labelled diagnostic, a keyword/flag router scored 0% on adversarial cases, whereas an intent-reasoning model router scored 100%, matching the oracle. Controlled mechanism probes revealed three failure-handling tiers: tool faults recovered fully (1.0), ambiguous delegation recovered partially (0.30), and three latent or semantic modes never recovered (0.0). Blogger's Review: OrchestraBench provides a powerful tool for evaluating the failure modes and recovery of multi-agent orchestration frameworks, helping developers optimize pipeline reliability and fault tolerance.