[MEDIUM] — Horizon HTTP fetches in `StellarTransactionsService` lack `AbortController` / request timeout, can block indefinitely on Horizon hangs

[Alqku]

Severity: Medium
Type: Bug
Scope: Stellar
Labels: bug, refactoring

Description

fetchTransactionWithRetry and fetchTransactionOperationsWithRetry (src/stellar/stellar-transactions.service.ts, lines ~95–170) use the global fetch API without an AbortController, signal, or explicit timeout. Horizon is a public best-effort service; a stalled TCP socket can keep the promise unresolved indefinitely. The withRetries(3, ...) wrapper only retries on rejection — there is no wall-clock cap.

A Horizon outage will exhaust Node's HTTP agent, eventually timing out adjacent gRPC Soroban polling and the stellar-event.service.ts SSE stream. The donation-create endpoint can therefore hang, and the connection pool fill up.

Recommendation

• Wrap every Horizon fetch in AbortSignal.timeout(ms) (e.g. 8s) and a per-request AbortController you can pass into the helper.
• Bound the total retry budget with a wall-clock cap (e.g. 25s).
• Surface a ServiceUnavailableException immediately on timeout instead of retrying on the same socket.

---

[ZuLu0890]

I would love work on this issue

[MarcusDavidG]

I can handle this issue and will like to work on it, Please assign thank you.