Documentation issue

The plugin hook documentation currently describes timeoutMs as aborting a hook handler. The implementation only stops awaiting the handler; the underlying promise and side effects continue.

That distinction is important for host adapters: once the hook timeout expires, Gateway suspension accounting can stop waiting even though an external scheduler write is still in progress.

The cron hook documentation also recommends an initial full-list reconciliation but does not provide a complete safe adapter pattern for coalescing overlapping changes, waiting for durable external acceptance, retrying, or cleanup.

Location

• docs/plugins/hooks.md
• Cron hook/lifecycle sections in the plugin SDK documentation

Expected documentation

• State precisely that timeoutMs bounds how long OpenClaw awaits a hook; it does not cancel the handler.
• Provide a vendor-neutral cron host-adapter example that:
  • reconciles the full list at the authoritative cron lifecycle boundary;
  • serializes and coalesces overlapping cron_changed events;
  • publishes the full snapshot rather than applying lossy deltas;
  • keeps the hook promise pending until the external store durably accepts the snapshot;
  • retries transient failures without a core-owned outbox;
  • cleans up on gateway_stop.
• Explain the ownership boundary: OpenClaw exposes lifecycle facts; the host/plugin owns external scheduler durability and policy.

Additional context

This should land after the typed cron reconciliation lifecycle contract so the example documents a real public API rather than a proposed one.