Backoff Configuration Bug in WorkerServiceClient

[aptgetgit]

Description

Duplicate line in WorkerServiceClient::new() where with_max_elapsed_time gets called twice, second call just overwrites the first. This sets max_elapsed_time to None and retries never stop:

let backoff = ExponentialBackoffBuilder::new()
    .with_initial_interval(Duration::from_millis(100))
    .with_max_elapsed_time(Some(Duration::from_secs(10)))
    .with_max_elapsed_time(None)  // overwrites previous value
    .build();

Breaks

Breaks the failover logic in wait_tasks():

async fn wait_tasks(&self, proof_id: String, ids: &[String]) -> anyhow::Result<()> {
    // Try to create a subscriber first, otherwise use the failover mechanism.
    match self.create_subscriber(proof_id.clone()).await {
        Ok((sub_tx, mut sub_rx)) => { /* use subscriber */ }
        Err(err) => {
            tracing::warn!("Failed to create subscriber, using failover mechanism");
            self.wait_tasks_failover(proof_id, ids).await
        }
    }
}

Inside create_subscriber(), the buggy backoff is used for open_sub:

let response = backoff::future::retry(backoff.clone(), || async {
    connection
        .clone()
        .open_sub(request.clone())
        .await
        .map_err(status_to_backoff_error)
})
.await?;  // blocks here indefinitely with max_elapsed_time: None

When open_sub hits transient errors, it should timeout and fall back to wait_tasks_failover(). But with infinite retries, the error branch never runs and the polling fallback never kicks in.

Fix

Just remove the duplicate line:

let backoff = ExponentialBackoffBuilder::new()
    .with_initial_interval(Duration::from_millis(100))
    .with_max_elapsed_time(Some(Duration::from_secs(10)))
    .build();