Cost and duration tracking appears stale during long-running tasks

[robotlearning123]

Problem

When monitoring tasks with cc-m list, the cost and duration fields appear to freeze during execution. Multiple checks over ~5 minutes showed identical values.

Observed Behavior

# Check 1 (t=0)
f3b31fb784884a86 running $0.29 73.9s
623b165a4c6f4e7d running $0.87 165.7s

# Check 2 (t+60s)
f3b31fb784884a86 running $0.29 73.9s  # same!
623b165a4c6f4e7d running $0.87 165.7s # same!

# Check 3 (t+120s)  
f3b31fb784884a86 running $0.29 73.9s  # still same
623b165a4c6f4e7d running $0.75 225.2s # jumped, then froze again

Expected Behavior

Cost and duration should update in real-time (or near-real-time) as the agent works.

Possible Cause

Cost/duration may only update when the agent CLI streams a new chunk, not on a timer. If the agent is waiting for a long pytest run, there are no intermediate updates.