Problem

Currently, messages in the delivery queue can retry indefinitely without a configurable upper limit. When Telegram/other channels are temporarily unavailable, failed items accumulate and may be re-delivered unexpectedly after gateway restarts.

Proposed Solution

Add a delivery.maxRetries configuration option:

{
  "delivery": {
    "maxRetries": 3,
    "failedDir": "~/.openclaw/delivery-queue/failed/"
  }
}

Behavior:

• After maxRetries attempts, move item to failedDir instead of retrying
• Items in failedDir are never auto-retried
• Emit a system event or log entry when an item is moved to failed

Current Workaround

Using a LaunchAgent script that monitors failed/ directory and sends Telegram alerts, plus Heartbeat scanning. But this is a workaround — native support would be cleaner.

Use Case

Prevents 'ghost message' delivery after extended outages, while still allowing inspection of failed items.