Scenario: Agent Retries a Blocked Payment

The retry-loop scenario: why a blocked payment must be terminal for the attempt.

The setup

Your agent attempts a payment to a wallet that screening blocks. The agent, designed to be resilient, retries — 10 times in 30 seconds.

Without controls

• Each retry is another attempt — and if ANY retry path bypasses the screen, it settles
• No log of the attempts
• One bypass = one violation

With controls

1. Every attempt hits the same screen — blocked, logged
2. Rate limiting stops the retry storm
3. The agent surfaces the block instead of looping
4. The audit log shows all N attempts with the same decision

The takeaway

A blocked payment must be terminal for the attempt — the agent reports, not retries. Fail-closed + rate limits make the retry loop harmless.