dekaf: Remove connection healthcheck as FindCoordinator takes ~3-5 …

…seconds to respond

crates/dekaf/src/api_client.rs

@@ -297,29 +297,10 @@ impl deadpool::managed::Manager for KafkaConnectionParams {
 
     async fn recycle(
         &self,
-        conn: &mut BoxedKafkaConnection,
+        _conn: &mut BoxedKafkaConnection,
         _: &deadpool::managed::Metrics,
     ) -> deadpool::managed::RecycleResult<anyhow::Error> {
-        // Other than auth (and `ApiVersions` apparently) Kafka connections themselves are stateless
-        // so the only thing we need to do when recycling a connection
-        // is to confirm that it's still connected.
-        send_request(
-            conn,
-            messages::FindCoordinatorRequest::default()
-                .with_key(protocol::StrBytes::from_string("test".to_string()))
-                .with_key_type(0),
-            Some(
-                messages::RequestHeader::default()
-                    .with_request_api_key(messages::FindCoordinatorRequest::KEY)
-                    .with_request_api_version(3),
-            ),
-        )
-        .await
-        .map(|_| ())
-        .map_err(|e| {
-            tracing::warn!(err=?e, broker=self.broker_url, "Connection failed healthcheck");
-            deadpool::managed::RecycleError::Backend(e)
-        })
+        Ok(())
     }
 }