Full logs + artifacts: https://gist.github.com/Gernspi/98ea2447e9502fff2ab64e89cdcdbecf

Titel: stale_running tasks remain in ledger and cannot be cleared in OpenClaw 2026.4.2

Setup:

• OpenClaw 2026.4.2
• user systemd service: openclaw-gateway.service
• no separate worker/runner systemd unit
• only gateway process visible besides the shell/session

Observed behavior:

• openclaw tasks audit reports many stale_running findings
• openclaw tasks maintenance --apply reports success but does not reduce findings
• restarting openclaw-gateway does not clear them
• openclaw tasks list --status running --json still shows many tasks as running

Important inconsistency:

• cancelling a listed subagent task returns: Subagent task not found.
• cancelling the paired cli task returns: Task runtime does not support cancellation yet.

Example pair:

• subagent taskId: 1d484e85-0a06-486e-8c29-e6682b14dbd3
• cli taskId: 1fac6326-b1c5-4eac-a5b7-a054f47c6a00

Expected:

• stale running tasks should either be cancellable, removable by maintenance, or reconciled after gateway restart

Actual:

• task ledger continues to report stale running tasks that are no longer backed by a real subagent process and cannot be cleared through the available CLI paths

Artifacts attached:

• version.txt
• tasks-audit.txt
• tasks-running.json

Original cancel outputs:

• subagent cancel: Subagent task not found.
• cli cancel: Task runtime does not support cancellation yet.

Local workaround that resolved the stuck ledger:

• stopped gateway
• backed up runs.sqlite / registry.sqlite / subagents/runs.json
• updated task_runs in ~/.openclaw/tasks/runs.sqlite from status='running' to status='failed' for stale entries
• restarted gateway

Result after workaround:

• 0 running tasks
• tasks audit: 0 errors, 2 warnings