fix(federation): reliable symmetric auto-federation across LAN/Tor/FIPS

Federated nodes failed to converge to full-mesh across the LAN<->Tailscale
boundary: nodes were invisible to peers, sync 'took ages'/timed out, and
names only updated on a manual sync. Onions were healthy in both directions
(~3-5s); the failures were app-layer.

- B: federation dials fast-fail a dead FIPS path via .fips_timeout(6s) in
  sync_with_peer + notify_join, so the Tor fallback isn't stuck behind the
  full 30s FIPS budget when LAN and remote peers share no FIPS path.
- A: notify_join (peer-joined) now spawns with retries+backoff instead of a
  single awaited best-effort POST, so the join RPC returns instantly (no
  'Request timeout') and the inviter reliably learns the joiner (was
  asymmetric).
- C: new 90s periodic federation auto-sync (none existed) so renamed nodes
  and roster changes propagate without a manual Sync click.
- self-heal: each auto-sync re-asserts membership to any peer that doesn't
  list us back, converging the fleet to full-mesh and healing pre-existing
  asymmetry with no manual re-joins.

Validated live across 7 nodes: a previously fleet-invisible node became
fully meshed automatically (logs: 'auto-sync ... reasserted=1',
'peer-joined ... delivered').

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

core/archipelago/src/federation/invites.rs

@@ -254,11 +254,57 @@ pub(crate) async fn notify_join(
         "params": params,
     });
 
-    let _ = crate::fips::dial::PeerRequest::new(remote_fips_npub, remote_onion, "/rpc/v1")
-        .service(crate::settings::transport::PeerService::Federation)
-        .timeout(std::time::Duration::from_secs(30))
-        .send_json(&body)
-        .await;
+    // Deliver the notification in the BACKGROUND with retries, and return
+    // immediately. Two reasons:
+    //   1. The join RPC must not block on this. Awaiting a cold FIPS overlay
+    //      (no shared FIPS path between LAN and remote/Tailscale peers) stalled
+    //      the whole join until FIPS timed out, surfacing as "Request timeout"
+    //      in the UI even though the local membership was already saved.
+    //   2. If this single best-effort POST failed, the inviter never learned
+    //      about us → asymmetric federation (they couldn't see us). Retrying in
+    //      the background until it lands makes federation converge to symmetric.
+    // `fips_timeout` fast-fails a dead FIPS path so the Tor fallback (which
+    // answers an onion in ~3-5s) is reached quickly on each attempt.
+    let remote_onion = remote_onion.to_string();
+    let remote_fips_npub = remote_fips_npub.map(|s| s.to_string());
+    tokio::spawn(async move {
+        // ~5 attempts with linear backoff: 0s, 10s, 20s, 30s, 40s — covers a
+        // peer that is briefly unreachable (restarting, publishing its onion)
+        // without hammering it.
+        for attempt in 1..=5u32 {
+            let res = crate::fips::dial::PeerRequest::new(
+                remote_fips_npub.as_deref(),
+                &remote_onion,
+                "/rpc/v1",
+            )
+            .service(crate::settings::transport::PeerService::Federation)
+            .timeout(std::time::Duration::from_secs(30))
+            .fips_timeout(std::time::Duration::from_secs(6))
+            .send_json(&body)
+            .await;
+            match res {
+                Ok((resp, transport)) if resp.status().is_success() => {
+                    tracing::info!(
+                        attempt,
+                        transport = %transport,
+                        "peer-joined notification delivered to inviter"
+                    );
+                    return;
+                }
+                Ok((resp, _)) => tracing::warn!(
+                    attempt,
+                    status = %resp.status(),
+                    "peer-joined notification rejected; will retry"
+                ),
+                Err(e) => tracing::warn!(attempt, error = %e, "peer-joined notification failed; will retry"),
+            }
+            tokio::time::sleep(std::time::Duration::from_secs(10 * attempt as u64)).await;
+        }
+        tracing::warn!(
+            onion = %remote_onion,
+            "peer-joined notification gave up after retries — peer may not see us until next sync"
+        );
+    });
     Ok(())
 }
 

core/archipelago/src/federation/mod.rs

@@ -12,6 +12,9 @@ mod types;
 
 // Re-export all public items so `crate::federation::*` continues to work.
 pub use invites::{accept_invite, create_invite};
+// Crate-internal: used by the periodic federation auto-sync to re-assert
+// membership to peers that don't list us back (asymmetry self-heal).
+pub(crate) use invites::notify_join;
 #[allow(unused_imports)]
 pub use storage::{
     add_node, fips_npub_for_onion, load_nodes, load_removed_dids, record_peer_transport,

core/archipelago/src/federation/sync.rs

@@ -33,6 +33,12 @@ pub async fn sync_with_peer(
         .header("X-Federation-Sig", signature)
         .header("X-Federation-Timestamp", timestamp)
         .timeout(std::time::Duration::from_secs(30))
+        // Fast-fail a cold/unreachable FIPS overlay (common between LAN and
+        // remote/Tailscale peers that share no FIPS path) so the Tor fallback —
+        // which answers an onion in ~3-5s — isn't stuck behind the full 30s FIPS
+        // budget. Without this, a state sync to a FIPS-unreachable peer "took
+        // ages" and join/sync appeared to time out even though Tor was healthy.
+        .fips_timeout(std::time::Duration::from_secs(6))
         .send_json(&body)
         .await
         .context("Failed to reach federated peer")?;