OpenClaw and Hermes are tightening the unattended-run paths after release day
The freshest May 8 work is less about new features and more about keeping agents from quietly failing while nobody is watching. OpenClaw now has fixes or reports for cron payload timeouts being collapsed back to a 120s idle watchdog, empty Heartbeat files still burning tokens, subagent completion fallback announcements, Nix-store plugin hardlinks, and fail-closed config writes. Hermes is seeing parallel hardening around cron lock scope and heartbeat ticks, Feishu gateway restarts replaying stale messages into restart loops, safer update installs, refusal to use destructive git reset recovery, terminal /doctor diagnostics, and native-Windows install/startup gaps.
The common thread is not one spectacular bug; it is the boring control plane that decides whether long-running agents keep doing the right thing after release day. Cron, Heartbeat, delivery fallback, update safety, and gateway replay behavior are where small regressions become expensive operator surprises.
- OpenClaw PR #79426 says explicit cron payload timeouts can be stripped when equal to the configured default, making the LLM idle watchdog fall back to 120s despite a longer per-run setting
- OpenClaw issue #79425 reports Heartbeat running every 30 minutes even when heartbeat.md contains no valid tasks, creating periodic token consumption
- OpenClaw PR #79405 hardens subagent completion fallback delivery so transient primary announce failures are not persisted as terminal before retry/give-up fallback can resolve
- OpenClaw PR #79344 allows Nix-store external plugin hardlinks only under OPENCLAW_NIX_MODE=1 while keeping normal hardlink rejection intact
- OpenClaw PR #79299 rejects unsafe best-effort config overwrites after unreadable existing config snapshots instead of writing tiny fallback-derived payloads
- Hermes PR #21901 narrows cron file-lock scope and adds a heartbeat ticker so long-running jobs do not block subsequent scheduling ticks
- Hermes PR #21902 skips stale Feishu pre-connect messages so gateway restarts do not replay old commands into restart loops
- Hermes PRs #20753 and #20759 keep update installs from mutating package-lock files or using destructive reset-hard recovery after fast-forward failure
- Hermes PR #21584 adds /doctor output for terminal and SSH clipboard failures; PR #21561 closes native-Windows install, UTF-8, tzdata, and startup gaps
- Most items are PR-stage or issue-stage, not new release tags
- The cluster matters most for operators who rely on cron, Heartbeat, subagents, Feishu, Nix packaging, native Windows, or remote terminals
- Because many fixes touch fallback and recovery paths, validate in staging instead of assuming a green startup proves unattended reliability