Finish the non-dialing send on master's own paths, and tidy what the merge left

The two node-side inline sends on the UDP encrypt-worker fallback paths
use send_existing. Both only ever see a UDP handle, for which
send_existing forwards to the same send, so behaviour is unchanged; it
leaves no node-side caller of the dialing send.

The medium-change heartbeat note said the send after a failed write
redials the peer. The link send no longer dials: once the stranded
connection is evicted, the next send fails at once and starts a
background connect, and only toward an address this node dialed.

The classification test lists NotConnected among the terminal errors,
so flipping it to transient fails a test that names the change. The
rx-stall tests drop the blocking read timeouts their polling read never
used, the Nym send_existing tests reuse the existing mock-proxy fixture,
and the SOCKS5 pool tests use the shared wait_until helper.

src/node/handlers/netmon.rs

@@ -167,11 +167,12 @@ impl Node {
     ///
     /// So a peer on TCP, Tor, Nym or BLE keeps the periodic heartbeat it had
     /// before this detector existed, and `link_dead_timeout_secs` remains the
-    /// backstop. Note that it does *not* recover by redialling: `send_async`
+    /// backstop. Note that it does *not* recover by redialling. The link send
-    /// only dials when the pool holds no connection for the address, and a
+    /// never dials, and a connection stranded by a medium change is still in
-    /// connection stranded by a medium change is still in the pool. It is
+    /// the pool until its writer's write to it fails. The send after that
-    /// evicted after a write to it fails, so the redial happens on the send
+    /// finds no connection and fails, starting a background connect only if
-    /// after the failure, not on the first one. Doing better for them means
+    /// this node dialed the peer at that address, and a later send uses it;
+    /// a peer that dialed in is never redialled. Doing better for them means
     /// dropping the stale connection
     /// rather than writing into it, which is a different change with a real
     /// cost behind it — a Tor peer pays a fresh circuit — and is not this one.

src/node/handlers/session.rs

@@ -2930,7 +2930,7 @@ impl Node {
             debug!(next_hop = %next_hop_addr, "Transport gone before inline send of session data");
             return;
         };
-        if let Err(error) = transport.send(remote_addr, &wire).await {
+        if let Err(error) = transport.send_existing(remote_addr, &wire).await {
             debug!(next_hop = %next_hop_addr, %error, "Inline send of session data failed");
         }
     }

src/node/mod.rs

@@ -3941,7 +3941,7 @@ impl Node {
                                 reason: format!("encryption failed: {}", e),
                             })?;
                             transport
-                                .send(&remote_addr, &wire)
+                                .send_existing(&remote_addr, &wire)
                                 .await
                                 .map_err(|e| link_send_error(*node_addr, e))?
                         }

src/node/tests/rx_stall.rs

@@ -148,11 +148,7 @@ async fn prime_link(node: &Node, bh: &Blackhole) -> std::net::TcpStream {
         tokio::time::sleep(Duration::from_millis(10)).await;
     }
     let (accepted, _) = bh.listener.accept().unwrap();
-    let accepted = std::net::TcpStream::from(accepted);
+    std::net::TcpStream::from(accepted)
-    accepted
-        .set_read_timeout(Some(Duration::from_millis(1000)))
-        .unwrap();
-    accepted
 }
 
 /// Close the node's connection to `bh` from the far end, after taking the
@@ -581,9 +577,6 @@ async fn msg1_resend_to_dead_outbound_leg_recovers_after_background_connect() {
     }
     let (accepted, _) = bh.listener.accept().unwrap();
     let mut accepted = std::net::TcpStream::from(accepted);
-    accepted
-        .set_read_timeout(Some(Duration::from_millis(1000)))
-        .unwrap();
     println!(
         "msg1 resend  sent after {:?} over the background connect",
         start.elapsed()

src/transport/mod.rs

@@ -1996,6 +1996,8 @@ mod tests {
             // answer, not this node's inability to transmit.
             TransportError::Timeout,
             TransportError::ConnectionRefused,
+            // The connection is gone and nothing will bring it back.
+            TransportError::NotConnected,
         ] {
             assert!(
                 !terminal.is_transient(),

src/transport/nym/mod.rs

@@ -1286,45 +1286,11 @@ mod tests {
         dest.stop_async().await.unwrap();
     }
 
-    /// A destination TCP transport behind a mock SOCKS5 proxy, and a
-    /// started Nym transport pointed at the proxy.
-    async fn nym_behind_mock_proxy() -> (
-        NymTransport,
-        TcpTransport,
-        crate::transport::PacketRx,
-        TransportAddr,
-    ) {
-        let (dest_tx, dest_rx) = packet_channel(32);
-        let dest_config = TcpConfig {
-            bind_addr: Some("127.0.0.1:0".to_string()),
-            ..Default::default()
-        };
-        let mut dest = TcpTransport::new(TransportId::new(100), None, dest_config, dest_tx);
-        dest.start_async().await.unwrap();
-        let dest_addr = dest.local_addr().unwrap();
-
-        let mock = MockSocks5Server::new(dest_addr).await.unwrap();
-        let proxy_addr = mock.addr();
-        let _proxy_handle = mock.spawn();
-
-        let (tx, _rx) = packet_channel(32);
-        let config = NymConfig {
-            socks5_addr: Some(proxy_addr.to_string()),
-            startup_timeout_secs: Some(5),
-            connect_timeout_ms: Some(5000),
-            ..Default::default()
-        };
-        let mut t = NymTransport::new(TransportId::new(200), None, config, tx);
-        t.start_async().await.unwrap();
-        let target = TransportAddr::from_string(&dest_addr.to_string());
-        (t, dest, dest_rx, target)
-    }
-
     /// With no pooled connection and no connect under way, `send_existing`
     /// fails with `NotConnected` and opens nothing.
     #[tokio::test]
     async fn send_existing_without_connection_fails_fast_and_dials_nothing() {
-        let (mut t, mut dest, _dest_rx, target) = nym_behind_mock_proxy().await;
+        let (mut dest, _dest_rx, mut t, target) = nym_via_mock_proxy().await;
 
         let result = t.send_existing(&target, &build_msg1_frame()).await;
 
@@ -1352,7 +1318,7 @@ mod tests {
     /// carries the send, with no second connection opened.
     #[tokio::test]
     async fn send_existing_promotes_a_finished_background_connect_and_sends_on_it() {
-        let (mut t, mut dest, mut dest_rx, target) = nym_behind_mock_proxy().await;
+        let (mut dest, mut dest_rx, mut t, target) = nym_via_mock_proxy().await;
 
         t.connect_async(&target).await.unwrap();
         let mut waited = 0;

src/transport/socks5/pool.rs

@@ -417,6 +417,7 @@ pub(crate) async fn proxied_receive_loop<S: ProxiedStats, M>(
 #[cfg(test)]
 mod tests {
     use super::*;
+    use crate::testutil::wait_until;
     use crate::transport::packet_channel;
     use crate::transport::stream::{next_conn_id, park_writer};
     use portable_atomic::{AtomicU64, Ordering};
@@ -463,20 +464,6 @@ mod tests {
         }
     }
 
-    /// Poll `f` every 10ms until it holds or `limit` elapses.
-    async fn wait_until<F: FnMut() -> bool>(mut f: F, limit: Duration) -> bool {
-        let deadline = Instant::now() + limit;
-        loop {
-            if f() {
-                return true;
-            }
-            if Instant::now() >= deadline {
-                return false;
-            }
-            tokio::time::sleep(Duration::from_millis(10)).await;
-        }
-    }
-
     /// A writer whose write fails must not remove a newer connection that has
     /// taken its address in the pool, nor run `on_remove` for it.
     #[tokio::test]