diff --git a/colly.go b/colly.go
index b4b96e61e..fce752fdc 100644
--- a/colly.go
+++ b/colly.go
@@ -1417,10 +1417,13 @@ func (c *Collector) Cookies(URL string) []*http.Cookie {
 }
 
 // Clone creates an exact copy of a Collector without callbacks.
-// HTTP backend, robots.txt cache and cookie jar are shared
-// between collectors.
+// The visited-URL store, robots.txt cache, cookie jar and HTTP transport
+// are shared between collectors so cookies and the connection pool flow
+// across them. The *http.Client and the callback-protecting lock are
+// independent so per-collector settings like AllowURLRevisit take effect
+// on the clone's redirects without leaking back into the parent.
 func (c *Collector) Clone() *Collector {
-	return &Collector{
+	clone := &Collector{
 		AllowedDomains:         c.AllowedDomains,
 		AllowURLRevisit:        c.AllowURLRevisit,
 		CacheDir:               c.CacheDir,
@@ -1441,7 +1444,6 @@ func (c *Collector) Clone() *Collector {
 		TraceHTTP:              c.TraceHTTP,
 		Context:                c.Context,
 		store:                  c.store,
-		backend:                c.backend,
 		debugger:               c.debugger,
 		Async:                  c.Async,
 		redirectHandler:        c.redirectHandler,
@@ -1449,12 +1451,33 @@ func (c *Collector) Clone() *Collector {
 		htmlCallbacks:          make([]*htmlCallbackContainer, 0, 8),
 		xmlCallbacks:           make([]*xmlCallbackContainer, 0, 8),
 		scrapedCallbacks:       make([]ScrapedCallback, 0, 8),
-		lock:                   c.lock,
+		lock:                   &sync.RWMutex{},
 		requestCallbacks:       make([]RequestCallback, 0, 8),
 		responseCallbacks:      make([]ResponseCallback, 0, 8),
 		robotsMap:              c.robotsMap,
 		wg:                     &sync.WaitGroup{},
 	}
+
+	// Independent httpBackend + http.Client so CheckRedirect closes over
+	// the clone, not the parent. Without this, the redirect handler would
+	// evaluate the parent's AllowURLRevisit and filters, so toggling those
+	// on the clone has no effect on redirects.
+	//
+	// Transport, cookie jar and Timeout are kept shared so connection
+	// pooling and session cookies continue to flow between collectors.
+	// LimitRules are copied by value (the slice header) — each *LimitRule
+	// is still shared because LimitRule.Init is idempotent.
+	clone.backend = &httpBackend{
+		LimitRules: append([]*LimitRule(nil), c.backend.LimitRules...),
+		lock:       &sync.RWMutex{},
+		Client: &http.Client{
+			Transport:     c.backend.Client.Transport,
+			Jar:           c.backend.Client.Jar,
+			Timeout:       c.backend.Client.Timeout,
+			CheckRedirect: clone.checkRedirectFunc(),
+		},
+	}
+	return clone
 }
 
 func (c *Collector) checkRedirectFunc() func(req *http.Request, via []*http.Request) error {
diff --git a/colly_test.go b/colly_test.go
index 4a8dda9c2..432f1c854 100644
--- a/colly_test.go
+++ b/colly_test.go
@@ -2236,3 +2236,52 @@ func TestLimitRuleClone(t *testing.T) {
 		t.Error("clone.Init() must not mutate the source's unexported state")
 	}
 }
+
+// TestCloneAllowURLRevisitIndependent verifies that AllowURLRevisit on a
+// cloned Collector takes effect during redirects, independently of the
+// parent's setting.
+//
+// Previously Clone() reused the parent's *http.Client, including its
+// CheckRedirect closure — and that closure captures the *parent*
+// Collector. As a result, the clone's CheckRedirect evaluated the
+// parent's AllowURLRevisit field, so toggling AllowURLRevisit on the
+// clone had no effect on redirects: visiting a redirect endpoint twice
+// failed with "already visited" even though the clone allowed revisits.
+func TestCloneAllowURLRevisitIndependent(t *testing.T) {
+	ts := newTestServer()
+	defer ts.Close()
+
+	parent := NewCollector()
+	clone := parent.Clone()
+	clone.AllowURLRevisit = true
+
+	redirectURL := ts.URL + "/redirect"
+	finalURL := ts.URL + "/redirected/"
+
+	requestCount := make(map[string]int)
+	clone.OnRequest(func(r *Request) {
+		requestCount[r.URL.String()]++
+	})
+	responseCount := make(map[string]int)
+	clone.OnResponse(func(r *Response) {
+		responseCount[r.Request.URL.String()]++
+	})
+
+	for i := 0; i < 2; i++ {
+		if err := clone.Visit(redirectURL); err != nil {
+			t.Fatalf("visit %d: unexpected error: %v", i+1, err)
+		}
+	}
+	if requestCount[redirectURL] != 2 {
+		t.Errorf("redirect URL visited %d times, want 2", requestCount[redirectURL])
+	}
+	if responseCount[finalURL] != 2 {
+		t.Errorf("final URL visited %d times, want 2", responseCount[finalURL])
+	}
+
+	// Parent should keep its original (default false) AllowURLRevisit and
+	// not gain revisit semantics just because the clone toggled it.
+	if parent.AllowURLRevisit {
+		t.Error("parent.AllowURLRevisit was mutated by setting it on the clone")
+	}
+}