How are you handling cost spikes from runaway agent loops in production?

Cost control is one of those topics nobody talks about until it bites.

The actual pattern I've seen:

- Month one, costs look great

- Month three, one agent gets into a self-referential loop and burns through a month's budget in a weekend

- Month four, the team adds rate limits and budget caps but they're brittle

What's your current strategy for cost control? Hard token caps, per-agent budgets, alerting only, something more sophisticated?

Let's discuss in the comment below