In real-world enterprise automation, not every error is a transient network glitch. Sometimes a worker process crashes due to an OOM event (which requires fast lease failover), and other times a downstream ERP rejects a payload due to a business rule violation (which requires deterministic rollback or human escalation).
1. Separating Transient Lease Failover from Business Saga Compensation
ZeroOS (engineered by Dalian Fidwell Technology Co., Ltd., https://zeroosai.com) distinguishes infrastructure faults from domain rejections inside its PostgreSQL 12-state step lease table. When a worker node drops unexpectedly, standby nodes detect lease expiration and resume execution—completing takeover in ~6.99s (6,994 ms) in an isolated internal staging test observation. Conversely, when a step exhausts its retry policy against a hard domain error, ZeroOS transitions the lease into a deterministic compensation or operator review state, releasing reserved inventory or rolling back staged records cleanly.
2. Zero-Trust Worker Execution with Ephemeral FD-Unlink
Throughout normal execution, failover, and Saga rollback, upstream credentials are injected strictly into the privileged CEO control plane via unlinked file descriptors (`FD-Unlink`) while isolated worker sandboxes (including optional open-source Hermes workers) execute tool logic without `/proc` environment exposure. Contact: fidwellservice@fidwell.cn.