Problem Statement

Currently, cron jobs have no automatic model fallback mechanism. If a job specifies a specific model (e.g., kimi-coding/k2p5) and that model hits a rate limit or quota, the job simply fails with consecutiveErrors++.

The Self-Heal Cron may retry failed jobs, but it retries with the same model, which often results in repeated failures during rate-limiting events.

Current Workarounds

Users must either:

1. Set all cron jobs to a single "safe" model (expensive)
2. Implement manual fallback logic inside every cron prompt (tedious, error-prone)
3. Accept periodic job failures during quota/rate-limit events

Proposed Solution

Add a fallbacks field to cron job payloads, similar to how agents already support model fallbacks:

{
  "payload": {
    "kind": "agentTurn",
    "message": "Do the thing...",
    "model": "kimi-coding/k2p5",
    "fallbacks": [
      "anthropic/claude-sonnet-4-6",
      "openai-codex/gpt-5.3-codex"
    ]
  }
}

Expected Behavior

1. Cron executes with primary model
2. If primary fails with rate-limit/quota error → automatic retry with fallback[0]
3. If fallback[0] fails → retry with fallback[1]
4. Only increment consecutiveErrors if all models exhausted
5. Log which model succeeded for observability

Use Case

We run 20+ cron jobs for routine tasks (health checks, digests, monitoring). We want:

• Primary: Cheap/fast models (MiniMax, Kimi) for cost efficiency
• Fallback: Premium models (Sonnet, Opus) when quotas are hit
• Result: Jobs complete reliably without manual intervention or cost bloat

Prior Art

Agent configs already support this:

"agents": {
  "defaults": {
    "model": {
      "primary": "anthropic/claude-sonnet-4-6",
      "fallbacks": ["openai-codex/gpt-5.3-codex", "kimi-coding/k2p5"]
    }
  }
}

Extending this pattern to cron jobs would provide consistency across the platform.

Additional Context

• Runtime: OpenClaw 2026.2.23
• Current cron count: 22 jobs
• Pain point: Rate limits on budget models cause cascading job failures during high-activity periods

Happy to provide more details or help test if this gets prioritized!