Resuming Task agents re-runs work instead of returning cached results

Description

When Task agents hit rate limits and return with an agentId, resuming them with the resume parameter causes them to re-run the entire research instead of returning the cached/completed results.

Steps to Reproduce

1. Launch Task agents (Explore type) for research
2. Agents complete their work (e.g., 14.2k tokens, 23.8k tokens)
3. Hit rate limit before results are returned to user
4. Wait for rate limit to reset
5. Resume agents using the provided agentId
6. Expected: Return the cached results from the completed work
7. Actual: Agents re-run the entire research, consuming additional tokens

Example

# First run - agents completed work
⏺ 2 Explore agents finished (ctrl+o to expand)
   ├─ Research product-category linking patterns · 8 tool uses · 14.2k tokens
   │  ⎿  Done
   └─ Research product options fetching patterns · 17 tool uses · 23.8k tokens
      ⎿  Done
  ⎿  You've hit your limit · resets 3pm (Asia/Seoul)

# After limit reset - resume caused re-run
<Task resume="aabfa1f">...</Task>
# Result: Agent ran again, consuming more tokens

Impact

• Users lose tokens due to duplicate work
• Poor user experience when waiting for rate limits to reset
• Research work that was already completed gets discarded

Expected Behavior

When resuming an agent that has already completed its work, it should return the cached results without re-running the research.