[CT-82] stellar.rs has no test distinguishing a Horizon timeout from a Horizon error response

[mftee]

Overview

contract/src/stellar.rs (1285 lines) is the largest file in the crate and, despite CT-41 adding tests for its zero-coverage state, likely still treats a network timeout and a genuine Horizon-returned error (e.g. tx failed, bad sequence number) the same way for retry purposes — only the former should be retried; the latter usually shouldn't be, since retrying a rejected transaction as-is will just fail again.

Tasks

• Distinguish timeout/network errors (retryable) from Horizon-returned business errors (not blindly retryable) in the retry logic
• Add a test asserting a Horizon 400-class error response does not trigger the same retry loop as a timeout

Acceptance Criteria

• A Horizon-rejected transaction is not retried identically to a network timeout

Notes for Contributors

Comment below to be assigned.