The Silent Death Of Workflow Runners During Issue

A develop workflow ran to completion only to vanish mid-issue-draft transition - leaving the runner dead, the status stuck in a stale loop, and the next local issue hanging in limbo. This isn’t just a bug; it’s a silent failure chain leaving teams guessing whether work was lost. Despite three completed issues and a live draft issued for local-6, no failure record formed, no machine_complete event triggered, and no meaningful artifact written. The root? A runtime disconnect between workflow stages and persistent state. The workflow-state.json locked on develop_starting while the agent logged step-by-step progress. No error output, no heartbeat, just silence. Monitoring tools kept reporting an old run, prolonging confusion. Fixing this demands tighter state synchronization at handoff, immediate failure signaling when runners drop, and clearer failure metadata. The next draft might be your last if you don’t catch this silent exit. Are you prepared for a task that disappears before it finishes?