diff --git a/openflared/internal/frpc/manager.go b/openflared/internal/frpc/manager.go
index d9bd3130..de84575f 100644
--- a/openflared/internal/frpc/manager.go
+++ b/openflared/internal/frpc/manager.go
@@ -156,28 +156,57 @@ func (m *Manager) restartProcess(ctx context.Context, relayID string, configPath
 	m.processes[relayID] = proc
 
 	go func() {
+		backoff := 1 * time.Second
+		const maxBackoff = 60 * time.Second
+
 		for {
-			select {
-			case <-procCtx.Done():
+			m.mu.Lock()
+			if procCtx.Err() != nil {
+				m.mu.Unlock()
 				return
-			default:
 			}
+			m.mu.Unlock()
 
 			cmd := exec.CommandContext(procCtx, m.cfg.FrpcPath, "-c", configPath)
+
+			m.mu.Lock()
 			proc.Cmd = cmd
 			proc.Status = "running"
+			m.mu.Unlock()
 
+			startedAt := time.Now()
 			err := cmd.Run()
+
+			m.mu.Lock()
+			if procCtx.Err() != nil {
+				proc.Status = "stopped"
+				m.mu.Unlock()
+				return
+			}
+
 			if err != nil {
-				if procCtx.Err() != nil {
-					return
-				}
 				proc.LastError = err.Error()
 				proc.Status = "error"
 				slog.Error("frpc process exited unexpectedly", "relay_id", relayID, "error", err)
-				time.Sleep(5 * time.Second) // backoff
 			} else {
 				proc.Status = "stopped"
+				proc.LastError = "exited unexpectedly with code 0"
+				slog.Warn("frpc process exited unexpectedly with code 0", "relay_id", relayID)
+			}
+			m.mu.Unlock()
+
+			if time.Since(startedAt) >= 10*time.Second {
+				backoff = 1 * time.Second
+			}
+
+			select {
+			case <-procCtx.Done():
+				return
+			case <-time.After(backoff):
+				backoff = backoff * 2
+				if backoff > maxBackoff {
+					backoff = maxBackoff
+				}
 			}
 		}
 	}()
diff --git a/openflared/internal/frpc/manager_test.go b/openflared/internal/frpc/manager_test.go
new file mode 100644
index 00000000..a5c7608e
--- /dev/null
+++ b/openflared/internal/frpc/manager_test.go
@@ -0,0 +1,267 @@
+package frpc
+
+import (
+	"context"
+	"fmt"
+	"os"
+	"path/filepath"
+	"strings"
+	"testing"
+	"time"
+
+	"openflare-flared/internal/config"
+	"openflare/service"
+)
+
+// Helper to write control file for the dummy script
+func writeControl(t *testing.T, dir string, exitCode int, delaySeconds int) {
+	controlPath := filepath.Join(dir, "control.txt")
+	content := fmt.Sprintf("%d %d\n", exitCode, delaySeconds)
+	err := os.WriteFile(controlPath, []byte(content), 0644)
+	if err != nil {
+		t.Fatalf("failed to write control file: %v", err)
+	}
+}
+
+// Setup a dummy executable script that reads control.txt to decide exit code and sleep duration
+func setupDummyScript(t *testing.T) (string, string) {
+	dir := t.TempDir()
+	scriptPath := filepath.Join(dir, "dummy_frpc")
+
+	// On macOS/Linux, we write a shell script
+	scriptContent := fmt.Sprintf(`#!/bin/sh
+control_file="%s/control.txt"
+EXIT_CODE=0
+DELAY=0
+if [ -f "$control_file" ]; then
+    read -r EXIT_CODE DELAY < "$control_file"
+fi
+if [ -n "$DELAY" ] && [ "$DELAY" -gt 0 ] 2>/dev/null; then
+    sleep "$DELAY"
+fi
+exit "${EXIT_CODE:-0}"
+`, dir)
+
+	err := os.WriteFile(scriptPath, []byte(scriptContent), 0755)
+	if err != nil {
+		t.Fatalf("failed to write dummy script: %v", err)
+	}
+
+	return scriptPath, dir
+}
+
+// Helper to poll for status to eliminate timing flakiness in tests
+func assertStatusEventually(t *testing.T, m *Manager, relayID string, expectedStatus string, timeout time.Duration) {
+	deadline := time.Now().Add(timeout)
+	for time.Now().Before(deadline) {
+		m.mu.RLock()
+		proc, ok := m.processes[relayID]
+		m.mu.RUnlock()
+		if ok && proc.Status == expectedStatus {
+			return
+		}
+		time.Sleep(50 * time.Millisecond)
+	}
+	m.mu.RLock()
+	proc, ok := m.processes[relayID]
+	var got string
+	var errStr string
+	if ok {
+		got = proc.Status
+		errStr = proc.LastError
+	} else {
+		got = "not_found"
+	}
+	m.mu.RUnlock()
+	t.Fatalf("expected status eventually %s, got %s (err: %s)", expectedStatus, got, errStr)
+}
+
+func TestStartProcessSuccess(t *testing.T) {
+	scriptPath, dir := setupDummyScript(t)
+	writeControl(t, dir, 0, 5) // exit code 0, sleep 5s
+
+	cfg := &config.Config{
+		ServerURL:   "http://localhost:8080",
+		TunnelToken: "test-token",
+		FrpcPath:    scriptPath,
+		DataDir:     dir,
+		StatePath:   filepath.Join(dir, "flared-state.json"),
+	}
+
+	m := NewManager(cfg)
+
+	newConfig := &service.FlaredTunnelConfigResponse{
+		Version:  "1",
+		Checksum: "sum1",
+		Relays: []service.FlaredRelayInfo{
+			{
+				RelayNodeID: "relay-1",
+				Address:     "127.0.0.1:7000",
+				AuthToken:   "auth-1",
+			},
+		},
+		Proxies: nil,
+	}
+
+	err := m.UpdateConfig(context.Background(), newConfig)
+	if err != nil {
+		t.Fatalf("failed to UpdateConfig: %v", err)
+	}
+
+	assertStatusEventually(t, m, "relay-1", "running", 4*time.Second)
+
+	m.mu.RLock()
+	proc := m.processes["relay-1"]
+	m.mu.RUnlock()
+
+	proc.Cancel()
+	assertStatusEventually(t, m, "relay-1", "stopped", 4*time.Second) // wait for clean stop
+}
+
+func TestStartProcessFailureAndBackoff(t *testing.T) {
+	dir := t.TempDir()
+	invalidScriptPath := filepath.Join(dir, "non_existent_frpc")
+
+	cfg := &config.Config{
+		ServerURL:   "http://localhost:8080",
+		TunnelToken: "test-token",
+		FrpcPath:    invalidScriptPath,
+		DataDir:     dir,
+		StatePath:   filepath.Join(dir, "flared-state.json"),
+	}
+
+	m := NewManager(cfg)
+	newConfig := &service.FlaredTunnelConfigResponse{
+		Version:  "1",
+		Checksum: "sum1",
+		Relays: []service.FlaredRelayInfo{
+			{
+				RelayNodeID: "relay-1",
+				Address:     "127.0.0.1:7000",
+				AuthToken:   "auth-1",
+			},
+		},
+		Proxies: nil,
+	}
+
+	_ = m.UpdateConfig(context.Background(), newConfig)
+
+	assertStatusEventually(t, m, "relay-1", "error", 4*time.Second)
+
+	// Correct the path to dummy script
+	scriptPath, _ := setupDummyScript(t)
+	writeControl(t, filepath.Dir(scriptPath), 0, 5)
+
+	m.mu.Lock()
+	m.cfg.FrpcPath = scriptPath
+	m.mu.Unlock()
+
+	// Wait for backoff retry (1s backoff)
+	assertStatusEventually(t, m, "relay-1", "running", 4*time.Second)
+
+	m.mu.RLock()
+	proc := m.processes["relay-1"]
+	m.mu.RUnlock()
+	proc.Cancel()
+}
+
+func TestUnexpectedExit0CPUProtection(t *testing.T) {
+	scriptPath, dir := setupDummyScript(t)
+	// Start with immediate exit code 0
+	writeControl(t, dir, 0, 0)
+
+	cfg := &config.Config{
+		ServerURL:   "http://localhost:8080",
+		TunnelToken: "test-token",
+		FrpcPath:    scriptPath,
+		DataDir:     dir,
+		StatePath:   filepath.Join(dir, "flared-state.json"),
+	}
+
+	m := NewManager(cfg)
+	newConfig := &service.FlaredTunnelConfigResponse{
+		Version:  "1",
+		Checksum: "sum1",
+		Relays: []service.FlaredRelayInfo{
+			{
+				RelayNodeID: "relay-1",
+				Address:     "127.0.0.1:7000",
+				AuthToken:   "auth-1",
+			},
+		},
+		Proxies: nil,
+	}
+
+	_ = m.UpdateConfig(context.Background(), newConfig)
+
+	assertStatusEventually(t, m, "relay-1", "stopped", 4*time.Second)
+
+	m.mu.RLock()
+	proc := m.processes["relay-1"]
+	if !strings.Contains(proc.LastError, "exited unexpectedly with code 0") {
+		t.Errorf("expected LastError to record exit status 0 warning, got %s", proc.LastError)
+	}
+	m.mu.RUnlock()
+
+	proc.Cancel()
+}
+
+func TestBackoffReset(t *testing.T) {
+	scriptPath, dir := setupDummyScript(t)
+	// Rapid exit code 1 to increase backoff
+	writeControl(t, dir, 1, 0)
+
+	cfg := &config.Config{
+		ServerURL:   "http://localhost:8080",
+		TunnelToken: "test-token",
+		FrpcPath:    scriptPath,
+		DataDir:     dir,
+		StatePath:   filepath.Join(dir, "flared-state.json"),
+	}
+
+	m := NewManager(cfg)
+	newConfig := &service.FlaredTunnelConfigResponse{
+		Version:  "1",
+		Checksum: "sum1",
+		Relays: []service.FlaredRelayInfo{
+			{
+				RelayNodeID: "relay-1",
+				Address:     "127.0.0.1:7000",
+				AuthToken:   "auth-1",
+			},
+		},
+		Proxies: nil,
+	}
+
+	_ = m.UpdateConfig(context.Background(), newConfig)
+
+	// Wait to crash
+	assertStatusEventually(t, m, "relay-1", "error", 4*time.Second)
+
+	// Now make it run successfully for 11 seconds (exit code 0, sleep 11s)
+	writeControl(t, dir, 0, 11)
+
+	// Wait for next retry to start running
+	assertStatusEventually(t, m, "relay-1", "running", 4*time.Second)
+
+	// Wait for process to run for 10.5 seconds to trigger backoff reset
+	time.Sleep(10500 * time.Millisecond)
+
+	// Now make it crash again (exit code 1, sleep 0s)
+	writeControl(t, dir, 1, 0)
+
+	// Wait for it to finish and crash
+	assertStatusEventually(t, m, "relay-1", "error", 4*time.Second)
+
+	// It crashed. Since it ran for > 10s, backoff should have been reset to 1s.
+	// We make it healthy again (exit code 0, sleep 5)
+	writeControl(t, dir, 0, 5)
+
+	// Wait 1.5 seconds. If backoff was reset to 1s, it should be running now.
+	assertStatusEventually(t, m, "relay-1", "running", 4*time.Second)
+
+	m.mu.RLock()
+	proc := m.processes["relay-1"]
+	m.mu.RUnlock()
+	proc.Cancel()
+}