feat(sync): Phase 3 - resilience and health tracking

- Add RelayHealthTracker with DashMap
- Implement exponential backoff (5s -> 1h max)
- Handle dead relays (24h failures -> daily retry)
- Add startup jitter to prevent thundering herd
- Add NGIT_SYNC_MAX_BACKOFF_SECS config

7 files changed, 630 insertions, 19 deletions

diff --git a/src/config.rs b/src/config.rs
index a2a27be..441a14d 100644
--- a/src/config.rs
+++ b/src/config.rs
@@ -87,6 +87,10 @@ pub struct Config {
     /// URL of relay to sync kind 30617 events from (optional, enables proactive sync)
     #[arg(long, env = "NGIT_SYNC_RELAY_URL")]
     pub sync_relay_url: Option<String>,
+
+    /// Maximum backoff time in seconds for sync relay reconnection (default: 3600 = 1 hour)
+    #[arg(long, env = "NGIT_SYNC_MAX_BACKOFF_SECS", default_value_t = 3600)]
+    pub sync_max_backoff_secs: u64,
 }
 
 impl Config {
@@ -143,6 +147,7 @@ impl Config {
             metrics_connection_per_ip_abuse_threshold: 10,
             metrics_top_n_repos: 10,
             sync_relay_url: None,
+            sync_max_backoff_secs: 3600,
         }
     }
 }
diff --git a/src/http/nip11.rs b/src/http/nip11.rs
index e9e1c25..22e5b22 100644
--- a/src/http/nip11.rs
+++ b/src/http/nip11.rs
@@ -106,6 +106,7 @@ mod tests {
             metrics_connection_per_ip_abuse_threshold: 10,
             metrics_top_n_repos: 10,
             sync_relay_url: None,
+            sync_max_backoff_secs: 3600,
         };
 
         let doc = RelayInformationDocument::from_config(&config);
@@ -141,6 +142,7 @@ mod tests {
             metrics_connection_per_ip_abuse_threshold: 10,
             metrics_top_n_repos: 10,
             sync_relay_url: None,
+            sync_max_backoff_secs: 3600,
         };
 
         let doc = RelayInformationDocument::from_config(&config);
diff --git a/src/main.rs b/src/main.rs
index 31e7cf6..9273afd 100644
--- a/src/main.rs
+++ b/src/main.rs
@@ -51,13 +51,14 @@ async fn main() -> Result<()> {
             config.domain
         );
 
-        // Start SyncManager for proactive sync (Phase 2: multi-relay support)
+        // Start SyncManager for proactive sync (Phase 2: multi-relay support, Phase 3: health tracking)
         // Even without initial sync_relay_url, SyncManager can discover relays from stored announcements
         let sync_manager = SyncManager::new(
             config.sync_relay_url.clone(),
             config.domain.clone(),
             relay_with_db.database.clone(),
             relay_with_db.write_policy.clone(),
+            &config,
         );
 
         if config.sync_relay_url.is_some() {
diff --git a/src/sync/connection.rs b/src/sync/connection.rs
index 76cc8e8..319cbbd 100644
--- a/src/sync/connection.rs
+++ b/src/sync/connection.rs
@@ -9,6 +9,12 @@
 //!   1. Layer 1: kinds 30617 + 30618 (announcements)
 //!   2. Layer 2: A/a tags for repository events
 //!   3. Layer 3: E/e tags for related events (PRs, Issues, etc.)
+//!
+//! ## Phase 3 Features
+//!
+//! - Health tracking with success/failure reporting
+//! - Exponential backoff with health-aware delays
+//! - Dead relay detection and minimal retry
 
 use std::sync::Arc;
 use std::time::Duration;
@@ -17,6 +23,7 @@ use nostr_sdk::prelude::*;
 use tokio::sync::mpsc;
 
 use super::filter::FilterService;
+use super::health::RelayHealthTracker;
 
 /// Event received from the sync connection
 #[derive(Debug, Clone)]
@@ -169,47 +176,96 @@ impl SyncConnection {
     }
 }
 
-/// Reconnect loop with exponential backoff
+/// Reconnect loop with health-aware exponential backoff
+///
+/// This function manages the connection lifecycle with health tracking:
+/// - Checks health state before attempting connections
+/// - Reports success/failure to the health tracker
+/// - Respects backoff delays from the health tracker
+/// - Handles dead relay detection (24h+ failures)
 ///
 /// # Arguments
 /// * `url` - The relay URL to connect to
 /// * `tx` - Channel sender for synced events
 /// * `filter_service` - FilterService for building subscriptions
 /// * `our_domain` - Our relay's domain (used to extract remote domain)
+/// * `health_tracker` - Health tracker for managing connection state
 pub async fn connect_with_retry(
     url: &str,
     tx: mpsc::Sender<SyncedEvent>,
     filter_service: Arc<FilterService>,
     _our_domain: &str,
+    health_tracker: Arc<RelayHealthTracker>,
 ) {
-    let mut backoff = Duration::from_secs(1);
-    let max_backoff = Duration::from_secs(60);
-
     // Extract remote domain from URL
     let remote_domain = extract_domain_from_url(url).unwrap_or_else(|| url.to_string());
 
     loop {
+        // Check if we should attempt connection based on health state
+        if !health_tracker.should_attempt_connection(url) {
+            // Wait for remaining backoff
+            if let Some(remaining) = health_tracker.get_remaining_backoff(url) {
+                tracing::debug!(
+                    "Relay {} in backoff, waiting {:?} before retry",
+                    url,
+                    remaining
+                );
+                tokio::time::sleep(remaining).await;
+                continue;
+            }
+        }
+
+        // Log current health state for dead relays
+        if health_tracker.is_dead(url) {
+            tracing::info!(
+                "Attempting reconnection to dead relay {} (daily retry)",
+                url
+            );
+        }
+
         match SyncConnection::new(url, filter_service.clone(), &remote_domain).await {
             Ok(conn) => {
-                backoff = Duration::from_secs(1); // Reset backoff on successful connection
+                // Record successful connection
+                health_tracker.record_success(url);
+                tracing::info!("Sync connection established to {}", url);
+
+                // Run the connection (this blocks until disconnection)
                 conn.run(tx.clone()).await;
+
+                // Connection ended - record as failure for reconnection backoff
+                // (The connection ending is considered a failure even if it worked for a while)
+                health_tracker.record_failure(url);
                 tracing::warn!("Sync connection to {} ended, will reconnect", url);
             }
             Err(e) => {
+                // Record connection failure
+                health_tracker.record_failure(url);
+
+                let failure_count = health_tracker.get_failure_count(url);
+                let state = health_tracker.get_state(url);
+
                 tracing::error!(
-                    "Failed to connect to sync relay {}: {} (retrying in {:?})",
+                    "Failed to connect to sync relay {} (attempt #{}, state: {}): {}",
                     url,
-                    e,
-                    backoff
+                    failure_count,
+                    state,
+                    e
                 );
             }
         }
 
-        // Wait before reconnecting
-        tokio::time::sleep(backoff).await;
+        // Get the backoff duration from health tracker
+        // If the health tracker has no backoff set (shouldn't happen), use a small default
+        let wait_duration = health_tracker
+            .get_remaining_backoff(url)
+            .unwrap_or(Duration::from_secs(5));
 
-        // Exponential backoff
-        backoff = std::cmp::min(backoff * 2, max_backoff);
+        tracing::debug!(
+            "Waiting {:?} before reconnecting to {}",
+            wait_duration,
+            url
+        );
+        tokio::time::sleep(wait_duration).await;
     }
 }
 
diff --git a/src/sync/health.rs b/src/sync/health.rs
new file mode 100644
index 0000000..51bd5ae
--- /dev/null
+++ b/src/sync/health.rs
@@ -0,0 +1,475 @@
+//! Relay Health Tracking for GRASP-02 Proactive Sync
+//!
+//! This module implements health tracking for relay connections, including:
+//! - Health state machine (Healthy -> Degraded -> Dead)
+//! - Exponential backoff with configurable max delay
+//! - Dead relay detection after 24h of continuous failures
+//!
+//! ## Health States
+//!
+//! - **Healthy**: Working connection, no recent failures
+//! - **Degraded**: Connection failed, retrying with backoff
+//! - **Dead**: 24h+ of continuous failures, minimal retry (once per day)
+
+use std::sync::Arc;
+use std::time::{Duration, Instant};
+
+use dashmap::DashMap;
+
+use crate::config::Config;
+
+/// Duration threshold before a relay is considered dead (24 hours)
+const DEAD_THRESHOLD_HOURS: u64 = 24;
+
+/// How often dead relays are retried (once per 24 hours)
+const DEAD_RETRY_INTERVAL_HOURS: u64 = 24;
+
+/// Default maximum backoff duration in seconds (1 hour)
+const DEFAULT_MAX_BACKOFF_SECS: u64 = 3600;
+
+/// Base backoff duration in seconds
+const BASE_BACKOFF_SECS: u64 = 5;
+
+/// Health state of a relay connection
+#[derive(Debug, Clone, Copy, PartialEq, Eq)]
+pub enum HealthState {
+    /// Working connection, no recent failures
+    Healthy,
+    /// Connection failed, retrying with exponential backoff
+    Degraded,
+    /// 24h+ of continuous failures, minimal retry
+    Dead,
+}
+
+impl std::fmt::Display for HealthState {
+    fn fmt(&self, f: &mut std::fmt::Formatter<'_>) -> std::fmt::Result {
+        match self {
+            HealthState::Healthy => write!(f, "healthy"),
+            HealthState::Degraded => write!(f, "degraded"),
+            HealthState::Dead => write!(f, "dead"),
+        }
+    }
+}
+
+/// Health information for a single relay
+#[derive(Debug, Clone)]
+pub struct RelayHealth {
+    /// Current health state
+    pub state: HealthState,
+    /// Number of consecutive connection failures
+    pub consecutive_failures: u32,
+    /// Time of the first failure in the current failure streak
+    pub first_failure_time: Option<Instant>,
+    /// Time of the last failure
+    pub last_failure_time: Option<Instant>,
+    /// Time of the last successful connection
+    pub last_success_time: Option<Instant>,
+    /// Next time a connection attempt should be made
+    pub next_retry_at: Option<Instant>,
+}
+
+impl Default for RelayHealth {
+    fn default() -> Self {
+        Self {
+            state: HealthState::Healthy,
+            consecutive_failures: 0,
+            first_failure_time: None,
+            last_failure_time: None,
+            last_success_time: None,
+            next_retry_at: None,
+        }
+    }
+}
+
+impl RelayHealth {
+    /// Create a new RelayHealth with healthy state
+    pub fn new() -> Self {
+        Self::default()
+    }
+}
+
+/// Thread-safe relay health tracker using DashMap
+#[derive(Debug)]
+pub struct RelayHealthTracker {
+    health: DashMap<String, RelayHealth>,
+    max_backoff_secs: u64,
+}
+
+impl RelayHealthTracker {
+    /// Create a new RelayHealthTracker
+    pub fn new(config: &Config) -> Self {
+        Self {
+            health: DashMap::new(),
+            max_backoff_secs: config.sync_max_backoff_secs,
+        }
+    }
+
+    /// Create a new RelayHealthTracker with default settings
+    pub fn with_defaults() -> Self {
+        Self {
+            health: DashMap::new(),
+            max_backoff_secs: DEFAULT_MAX_BACKOFF_SECS,
+        }
+    }
+
+    /// Create a new RelayHealthTracker with custom max backoff
+    pub fn with_max_backoff(max_backoff_secs: u64) -> Self {
+        Self {
+            health: DashMap::new(),
+            max_backoff_secs,
+        }
+    }
+
+    /// Record a successful connection to a relay
+    ///
+    /// Resets the relay to Healthy state and clears failure counters.
+    pub fn record_success(&self, relay_url: &str) {
+        let now = Instant::now();
+        let mut entry = self.health.entry(relay_url.to_string()).or_default();
+        let health = entry.value_mut();
+
+        let old_state = health.state;
+
+        // Reset to healthy state
+        health.state = HealthState::Healthy;
+        health.consecutive_failures = 0;
+        health.first_failure_time = None;
+        health.last_failure_time = None;
+        health.last_success_time = Some(now);
+        health.next_retry_at = None;
+
+        if old_state != HealthState::Healthy {
+            tracing::info!(
+                "Relay {} recovered to healthy (was {:?})",
+                relay_url,
+                old_state
+            );
+        }
+    }
+
+    /// Record a connection failure for a relay
+    ///
+    /// Increments failure counter, updates state, and calculates next retry time.
+    pub fn record_failure(&self, relay_url: &str) {
+        let now = Instant::now();
+        let mut entry = self.health.entry(relay_url.to_string()).or_default();
+        let health = entry.value_mut();
+
+        let old_state = health.state;
+
+        // Set first_failure_time if this is a new failure streak
+        if health.first_failure_time.is_none() {
+            health.first_failure_time = Some(now);
+        }
+
+        health.consecutive_failures = health.consecutive_failures.saturating_add(1);
+        health.last_failure_time = Some(now);
+
+        // Check if we should transition to Dead state
+        if let Some(first_failure) = health.first_failure_time {
+            let failure_duration = now.duration_since(first_failure);
+            let dead_threshold = Duration::from_secs(DEAD_THRESHOLD_HOURS * 3600);
+
+            if failure_duration >= dead_threshold {
+                health.state = HealthState::Dead;
+                // Dead relays retry once per day
+                health.next_retry_at =
+                    Some(now + Duration::from_secs(DEAD_RETRY_INTERVAL_HOURS * 3600));
+
+                if old_state != HealthState::Dead {
+                    tracing::warn!(
+                        "Relay {} marked dead after 24h failures ({} consecutive failures)",
+                        relay_url,
+                        health.consecutive_failures
+                    );
+                }
+            } else {
+                // Degraded state with exponential backoff
+                health.state = HealthState::Degraded;
+                let backoff = Self::get_backoff_duration(
+                    health.consecutive_failures,
+                    self.max_backoff_secs,
+                );
+                health.next_retry_at = Some(now + backoff);
+
+                if old_state != HealthState::Degraded {
+                    tracing::warn!(
+                        "Relay {} degraded, backoff {:?}",
+                        relay_url,
+                        backoff
+                    );
+                } else {
+                    tracing::debug!(
+                        "Relay {} failure #{}, backoff {:?}",
+                        relay_url,
+                        health.consecutive_failures,
+                        backoff
+                    );
+                }
+            }
+        }
+    }
+
+    /// Check if a connection attempt should be made to a relay
+    ///
+    /// Returns true if:
+    /// - The relay has no health record (first attempt)
+    /// - The relay is healthy
+    /// - The backoff period has elapsed
+    pub fn should_attempt_connection(&self, relay_url: &str) -> bool {
+        let entry = self.health.get(relay_url);
+
+        match entry {
+            None => true, // No record, allow first attempt
+            Some(entry) => {
+                let health = entry.value();
+
+                match health.state {
+                    HealthState::Healthy => true,
+                    HealthState::Degraded | HealthState::Dead => {
+                        // Check if backoff period has elapsed
+                        match health.next_retry_at {
+                            None => true,
+                            Some(next_retry) => Instant::now() >= next_retry,
+                        }
+                    }
+                }
+            }
+        }
+    }
+
+    /// Get the current health state of a relay
+    pub fn get_state(&self, relay_url: &str) -> HealthState {
+        self.health
+            .get(relay_url)
+            .map(|entry| entry.value().state)
+            .unwrap_or(HealthState::Healthy)
+    }
+
+    /// Check if a relay is marked as dead
+    pub fn is_dead(&self, relay_url: &str) -> bool {
+        self.get_state(relay_url) == HealthState::Dead
+    }
+
+    /// Get the remaining backoff duration for a relay
+    ///
+    /// Returns None if no backoff is active.
+    pub fn get_remaining_backoff(&self, relay_url: &str) -> Option<Duration> {
+        let entry = self.health.get(relay_url)?;
+        let health = entry.value();
+        let next_retry = health.next_retry_at?;
+        let now = Instant::now();
+
+        if now >= next_retry {
+            None
+        } else {
+            Some(next_retry - now)
+        }
+    }
+
+    /// Get the consecutive failure count for a relay
+    pub fn get_failure_count(&self, relay_url: &str) -> u32 {
+        self.health
+            .get(relay_url)
+            .map(|entry| entry.value().consecutive_failures)
+            .unwrap_or(0)
+    }
+
+    /// Calculate the backoff duration based on failure count
+    ///
+    /// Uses exponential backoff: base * 2^failures, capped at max_backoff
+    pub fn get_backoff_duration(consecutive_failures: u32, max_backoff_secs: u64) -> Duration {
+        let backoff_secs = BASE_BACKOFF_SECS
+            .saturating_mul(2u64.saturating_pow(consecutive_failures.saturating_sub(1)));
+        Duration::from_secs(backoff_secs.min(max_backoff_secs))
+    }
+
+    /// Get all tracked relay URLs
+    pub fn get_tracked_relays(&self) -> Vec<String> {
+        self.health.iter().map(|entry| entry.key().clone()).collect()
+    }
+
+    /// Get a clone of the health info for a relay
+    pub fn get_health(&self, relay_url: &str) -> Option<RelayHealth> {
+        self.health.get(relay_url).map(|entry| entry.value().clone())
+    }
+}
+
+/// Create a shared RelayHealthTracker wrapped in Arc
+pub fn create_health_tracker(config: &Config) -> Arc<RelayHealthTracker> {
+    Arc::new(RelayHealthTracker::new(config))
+}
+
+#[cfg(test)]
+mod tests {
+    use super::*;
+
+    #[test]
+    fn test_health_state_display() {
+        assert_eq!(HealthState::Healthy.to_string(), "healthy");
+        assert_eq!(HealthState::Degraded.to_string(), "degraded");
+        assert_eq!(HealthState::Dead.to_string(), "dead");
+    }
+
+    #[test]
+    fn test_default_health_is_healthy() {
+        let health = RelayHealth::default();
+        assert_eq!(health.state, HealthState::Healthy);
+        assert_eq!(health.consecutive_failures, 0);
+        assert!(health.first_failure_time.is_none());
+    }
+
+    #[test]
+    fn test_should_attempt_connection_new_relay() {
+        let tracker = RelayHealthTracker::with_defaults();
+        assert!(tracker.should_attempt_connection("wss://new-relay.example.com"));
+    }
+
+    #[test]
+    fn test_record_success_resets_to_healthy() {
+        let tracker = RelayHealthTracker::with_defaults();
+        let url = "wss://test-relay.example.com";
+
+        // Simulate a few failures
+        tracker.record_failure(url);
+        tracker.record_failure(url);
+        assert_eq!(tracker.get_state(url), HealthState::Degraded);
+        assert_eq!(tracker.get_failure_count(url), 2);
+
+        // Record success
+        tracker.record_success(url);
+        assert_eq!(tracker.get_state(url), HealthState::Healthy);
+        assert_eq!(tracker.get_failure_count(url), 0);
+        assert!(tracker.should_attempt_connection(url));
+    }
+
+    #[test]
+    fn test_backoff_increases_exponentially() {
+        // failure 1: 5s
+        assert_eq!(
+            RelayHealthTracker::get_backoff_duration(1, 3600),
+            Duration::from_secs(5)
+        );
+        // failure 2: 10s
+        assert_eq!(
+            RelayHealthTracker::get_backoff_duration(2, 3600),
+            Duration::from_secs(10)
+        );
+        // failure 3: 20s
+        assert_eq!(
+            RelayHealthTracker::get_backoff_duration(3, 3600),
+            Duration::from_secs(20)
+        );
+        // failure 4: 40s
+        assert_eq!(
+            RelayHealthTracker::get_backoff_duration(4, 3600),
+            Duration::from_secs(40)
+        );
+        // failure 5: 80s
+        assert_eq!(
+            RelayHealthTracker::get_backoff_duration(5, 3600),
+            Duration::from_secs(80)
+        );
+    }
+
+    #[test]
+    fn test_backoff_capped_at_max() {
+        let max_backoff = 3600u64;
+        // After many failures, should cap at max_backoff (1 hour)
+        assert_eq!(
+            RelayHealthTracker::get_backoff_duration(20, max_backoff),
+            Duration::from_secs(max_backoff)
+        );
+    }
+
+    #[test]
+    fn test_degraded_state_after_failure() {
+        let tracker = RelayHealthTracker::with_defaults();
+        let url = "wss://test-relay.example.com";
+
+        tracker.record_failure(url);
+        assert_eq!(tracker.get_state(url), HealthState::Degraded);
+        assert_eq!(tracker.get_failure_count(url), 1);
+    }
+
+    #[test]
+    fn test_backoff_blocks_immediate_reconnection() {
+        let tracker = RelayHealthTracker::with_defaults();
+        let url = "wss://test-relay.example.com";
+
+        tracker.record_failure(url);
+
+        // Immediately after failure, should not attempt (backoff active)
+        assert!(!tracker.should_attempt_connection(url));
+
+        // Remaining backoff should be some positive duration
+        let remaining = tracker.get_remaining_backoff(url);
+        assert!(remaining.is_some());
+        assert!(remaining.unwrap() > Duration::ZERO);
+    }
+
+    #[test]
+    fn test_is_dead() {
+        let tracker = RelayHealthTracker::with_defaults();
+        let url = "wss://test-relay.example.com";
+
+        // Initially not dead
+        assert!(!tracker.is_dead(url));
+
+        // After a failure, still not dead (just degraded)
+        tracker.record_failure(url);
+        assert!(!tracker.is_dead(url));
+        assert_eq!(tracker.get_state(url), HealthState::Degraded);
+    }
+
+    #[test]
+    fn test_get_tracked_relays() {
+        let tracker = RelayHealthTracker::with_defaults();
+
+        tracker.record_success("wss://relay1.example.com");
+        tracker.record_failure("wss://relay2.example.com");
+
+        let tracked = tracker.get_tracked_relays();
+        assert_eq!(tracked.len(), 2);
+        assert!(tracked.contains(&"wss://relay1.example.com".to_string()));
+        assert!(tracked.contains(&"wss://relay2.example.com".to_string()));
+    }
+
+    #[test]
+    fn test_custom_max_backoff() {
+        let custom_max = 60u64; // 1 minute max
+        let tracker = RelayHealthTracker::with_max_backoff(custom_max);
+        let url = "wss://test-relay.example.com";
+
+        // Simulate many failures
+        for _ in 0..20 {
+            tracker.record_failure(url);
+        }
+
+        // The remaining backoff should respect the custom max
+        // Note: We can't easily test the internal backoff calculation here,
+        // but we can verify the tracker was created with the custom setting
+        assert_eq!(tracker.max_backoff_secs, custom_max);
+    }
+
+    #[test]
+    fn test_get_health_returns_clone() {
+        let tracker = RelayHealthTracker::with_defaults();
+        let url = "wss://test-relay.example.com";
+
+        tracker.record_success(url);
+        let health = tracker.get_health(url);
+
+        assert!(health.is_some());
+        let health = health.unwrap();
+        assert_eq!(health.state, HealthState::Healthy);
+        assert!(health.last_success_time.is_some());
+    }
+
+    #[test]
+    fn test_get_health_nonexistent() {
+        let tracker = RelayHealthTracker::with_defaults();
+        let health = tracker.get_health("wss://nonexistent.example.com");
+        assert!(health.is_none());
+    }
+}
\ No newline at end of file
diff --git a/src/sync/manager.rs b/src/sync/manager.rs
index 8f6a9bd..1f70f42 100644
--- a/src/sync/manager.rs
+++ b/src/sync/manager.rs
@@ -9,18 +9,31 @@
 //! - Relay discovery from stored kind 30617 announcements
 //! - Multiple simultaneous relay connections
 //! - Three-layer filter strategy via FilterService
+//!
+//! ## Phase 3 Features
+//!
+//! - Health tracking with exponential backoff
+//! - Dead relay detection after 24h of failures
+//! - Startup jitter to prevent thundering herd
 
 use std::collections::HashSet;
 use std::sync::Arc;
+use std::time::Duration;
 
 use nostr_relay_builder::prelude::*;
+use rand::Rng;
 use tokio::sync::mpsc;
 
 use super::connection::{connect_with_retry, SyncedEvent};
 use super::filter::FilterService;
+use super::health::RelayHealthTracker;
 use super::SYNC_SOURCE_ADDR;
+use crate::config::Config;
 use crate::nostr::builder::{Nip34WritePolicy, SharedDatabase};
 
+/// Maximum startup jitter in milliseconds (10 seconds)
+const MAX_STARTUP_JITTER_MS: u64 = 10_000;
+
 /// Coordinates proactive sync from configured and discovered relays
 pub struct SyncManager {
     /// Initial relay URL to sync from (from config)
@@ -31,6 +44,8 @@ pub struct SyncManager {
     database: SharedDatabase,
     /// Write policy for validating events
     write_policy: Nip34WritePolicy,
+    /// Health tracker for relay connections
+    health_tracker: Arc<RelayHealthTracker>,
 }
 
 impl SyncManager {
@@ -41,17 +56,20 @@ impl SyncManager {
     /// * `relay_domain` - Our relay's domain (used to exclude self from sync)
     /// * `database` - Shared database for storing events and querying announcements
     /// * `write_policy` - Write policy for validating synced events
+    /// * `config` - Configuration for health tracking settings
     pub fn new(
         initial_relay_url: Option<String>,
         relay_domain: String,
         database: SharedDatabase,
         write_policy: Nip34WritePolicy,
+        config: &Config,
     ) -> Self {
         Self {
             initial_relay_url,
             relay_domain,
             database,
             write_policy,
+            health_tracker: Arc::new(RelayHealthTracker::new(config)),
         }
     }
 
@@ -68,9 +86,15 @@ impl SyncManager {
             relay_domain,
             database,
             write_policy,
+            health_tracker: Arc::new(RelayHealthTracker::with_defaults()),
         }
     }
 
+    /// Get a reference to the health tracker
+    pub fn health_tracker(&self) -> Arc<RelayHealthTracker> {
+        self.health_tracker.clone()
+    }
+
     /// Run the sync manager
     ///
     /// This discovers relays from stored announcements, spawns connection tasks,
@@ -94,12 +118,14 @@ impl SyncManager {
         // Track active relay URLs to avoid duplicates
         let mut active_relays: HashSet<String> = HashSet::new();
 
+        // Collect all relays to connect to
+        let mut relays_to_connect: Vec<String> = Vec::new();
+
         // Start with initial relay if configured
         if let Some(ref url) = self.initial_relay_url {
             if !self.is_own_relay(url) {
-                tracing::info!("Connecting to initial sync relay: {}", url);
+                relays_to_connect.push(url.clone());
                 active_relays.insert(url.clone());
-                self.spawn_connection(url.clone(), tx.clone(), filter_service.clone());
             } else {
                 tracing::info!("Skipping initial relay (is our own relay): {}", url);
             }
@@ -109,12 +135,17 @@ impl SyncManager {
         let discovered_urls = filter_service.discover_relay_urls().await;
         for url in discovered_urls {
             if !active_relays.contains(&url) && !self.is_own_relay(&url) {
-                tracing::info!("Connecting to discovered relay: {}", url);
+                relays_to_connect.push(url.clone());
                 active_relays.insert(url.clone());
-                self.spawn_connection(url, tx.clone(), filter_service.clone());
             }
         }
 
+        // Spawn connections with startup jitter to prevent thundering herd
+        for url in relays_to_connect {
+            tracing::info!("Scheduling connection to sync relay: {}", url);
+            self.spawn_connection_with_jitter(url, tx.clone(), filter_service.clone());
+        }
+
         if active_relays.is_empty() {
             tracing::warn!("No sync relays configured or discovered, SyncManager idle");
         } else {
@@ -133,6 +164,7 @@ impl SyncManager {
                 if !active_relays.contains(&url) && !self.is_own_relay(&url) {
                     tracing::info!("Discovered new relay from event, connecting: {}", url);
                     active_relays.insert(url.clone());
+                    // New relays discovered during runtime don't need jitter
                     self.spawn_connection(url, tx.clone(), filter_service.clone());
                 }
             }
@@ -148,7 +180,36 @@ impl SyncManager {
         url.contains(&self.relay_domain)
     }
 
-    /// Spawn a connection task for a relay
+    /// Spawn a connection task for a relay with startup jitter
+    ///
+    /// Adds a random delay (0-10s) before connecting to prevent thundering herd
+    /// on startup when multiple relays are configured.
+    fn spawn_connection_with_jitter(
+        &self,
+        url: String,
+        tx: mpsc::Sender<SyncedEvent>,
+        filter_service: Arc<FilterService>,
+    ) {
+        let domain = self.relay_domain.clone();
+        let health_tracker = self.health_tracker.clone();
+
+        tokio::spawn(async move {
+            // Apply startup jitter
+            let jitter_ms = rand::thread_rng().gen_range(0..MAX_STARTUP_JITTER_MS);
+            tracing::debug!(
+                "Applying {}ms startup jitter before connecting to {}",
+                jitter_ms,
+                url
+            );
+            tokio::time::sleep(Duration::from_millis(jitter_ms)).await;
+
+            connect_with_retry(&url, tx, filter_service, &domain, health_tracker).await;
+        });
+    }
+
+    /// Spawn a connection task for a relay without jitter
+    ///
+    /// Used for relays discovered during runtime (not at startup).
     fn spawn_connection(
         &self,
         url: String,
@@ -156,8 +217,10 @@ impl SyncManager {
         filter_service: Arc<FilterService>,
     ) {
         let domain = self.relay_domain.clone();
+        let health_tracker = self.health_tracker.clone();
+
         tokio::spawn(async move {
-            connect_with_retry(&url, tx, filter_service, &domain).await;
+            connect_with_retry(&url, tx, filter_service, &domain, health_tracker).await;
         });
     }
 
diff --git a/src/sync/mod.rs b/src/sync/mod.rs
index 1155eaf..653aa27 100644
--- a/src/sync/mod.rs
+++ b/src/sync/mod.rs
@@ -9,12 +9,21 @@
 //! - **Layer 1**: Announcement discovery (kinds 30617 + 30618)
 //! - **Layer 2**: Repository events (A/a tags for shared repos)
 //! - **Layer 3**: Related events (E/e tags for discussions, reviews)
+//!
+//! ## Resilience & Health Tracking (Phase 3)
+//!
+//! - **Health tracking**: Per-relay connection health states (Healthy, Degraded, Dead)
+//! - **Exponential backoff**: Smart retry delays on failures (5s -> 1h max)
+//! - **Dead relay handling**: Minimal retry for 24h+ failed relays
+//! - **Startup jitter**: Prevent thundering herd on launch (0-10s random delay)
 
 mod connection;
 mod filter;
+pub mod health;
 mod manager;
 
 pub use filter::FilterService;
+pub use health::{HealthState, RelayHealth, RelayHealthTracker};
 pub use manager::SyncManager;
 
 use std::net::SocketAddr;