Describe the bug

WhatsApp Web credentials file (creds.json) gets corrupted approximately every 35 minutes, which coincides with WhatsApp's pre-key rotation cycle. The system automatically restores from backup, but this causes intermittent connection failures and breaks scheduled task notifications.

Environment

• OpenClaw Version: 2026.4.1
• Node.js: v22.22.2
• OS: Linux (VPS, virtual disk /dev/vda3)
• Channel: WhatsApp Web
• Account: [REDACTED - Phone number removed for privacy]

Logs

12:39:08 WARN  restored corrupted WhatsApp creds.json from backup
13:14:08 WARN  restored corrupted WhatsApp creds.json from backup  
13:49:09 WARN  restored corrupted WhatsApp creds.json from backup

Corruption interval: ~35 minutes (matches pre-key rotation)

Impact

• Scheduled cron jobs fail to send notifications (3 tasks affected)
• Requires manual openclaw channels login to reconnect
• Gateway status 499 disconnections

Expected behavior

Credentials should be written atomically (write to temp file, then rename) to prevent corruption during pre-key rotation.

Possible cause

Race condition or non-atomic write in WhatsApp credential storage logic. The corruption occurs during the pre-key rotation write operation.

Workaround

Deployed monitoring script with:

• Auto-relogin on disconnect (every 2 min check)
• Automatic credential backup (every 30 min)
• Removed forced 30-min refresh (triggers rate limiting)

Additional context

After deploying the workaround, observed a secondary issue: WhatsApp gateway disconnects every ~60 seconds (status 499), possibly due to rate limiting from forced refresh. This was resolved by removing the 30-minute forced login task.

Reproduction steps:

1. Link WhatsApp Web account: openclaw channels login --channel whatsapp --account default
2. Wait ~35 minutes for pre-key rotation
3. Check logs for restored corrupted WhatsApp creds.json from backup
4. Observe connection drops and failed scheduled tasks