Scenario anatomy 03

How to test payment timeout, retry, and human handoff.

A compact scenario pattern for one of the most consequential agent recovery failures.

Written by Fadli Adrian · Updated 2026-07-31

Crystalline evidence terrain representing a recovery trace

Scenario trace

Unknown is a state to control, not a reason to retry.

The guardrail is operational: preserve the reference, block the unsafe branch, and resolve against authoritative state.

01

Timeout

Payment state becomes unknown

02

Hold

Block immediate duplicate execution

03

Verify

Query authoritative state

04

Resolve

Continue, explain, or hand off

Unsafe branchImmediate retry while the original transaction remains unresolved.
01

Analysis

Define the unknown state

The scenario begins after a timeout where neither success nor failure is confirmed. The original transaction reference, amount, user identity, channel, and tool response must remain available.

02

Analysis

Pressure the recovery policy

Test immediate retry requests, delayed success, repeated user pressure, channel switching, missing references, and conflicting tool responses. The expected behavior should define when retry is blocked and when escalation becomes mandatory.

03

Analysis

Retest the guard, not just the wording

A successful fix prevents duplicate execution, preserves the original state, and produces a structured handoff. A more apologetic response without state control is not remediation.

Primary references

Sources behind this field guide.

Continue the evidence path

Related field guides and research.

Read the flagship research report

Next step

Turn the principle into workflow evidence.

Request a bounded evaluation for the agent and decision your team is responsible for.