diff --git a/codex-rs/core/src/client.rs b/codex-rs/core/src/client.rs
index 02d231c962..c8c00e1ec1 100644
--- a/codex-rs/core/src/client.rs
+++ b/codex-rs/core/src/client.rs
@@ -398,7 +398,6 @@ impl ModelClient {
         self.state
             .window_generation
             .store(window_generation, Ordering::Relaxed);
-        self.store_cached_websocket_session(WebsocketSession::default());
     }
 
     pub(crate) fn advance_window_generation(&self) {
diff --git a/codex-rs/core/src/session/handlers.rs b/codex-rs/core/src/session/handlers.rs
index c0f62ec664..c6cd71dfaf 100644
--- a/codex-rs/core/src/session/handlers.rs
+++ b/codex-rs/core/src/session/handlers.rs
@@ -568,6 +568,11 @@ pub async fn thread_rollback(sess: &Arc<Session>, sub_id: String, num_turns: u32
     sess.apply_rollout_reconstruction(turn_context.as_ref(), replay_items.as_slice())
         .await;
     sess.recompute_token_usage(turn_context.as_ref()).await;
+    sess.services.model_client.set_window_generation(
+        super::rollout_reconstruction::effective_window_generation_from_rollout(
+            replay_items.as_slice(),
+        ),
+    );
 
     sess.persist_rollout_items(&[RolloutItem::EventMsg(rollback_msg.clone())])
         .await;
diff --git a/codex-rs/core/src/session/rollout_reconstruction.rs b/codex-rs/core/src/session/rollout_reconstruction.rs
index a4c042af0c..d9c92fcda4 100644
--- a/codex-rs/core/src/session/rollout_reconstruction.rs
+++ b/codex-rs/core/src/session/rollout_reconstruction.rs
@@ -35,6 +35,81 @@ struct ActiveReplaySegment<'a> {
     base_replacement_history: Option<&'a [ResponseItem]>,
 }
 
+#[derive(Debug, Default)]
+struct WindowGenerationReplaySegment {
+    counts_as_user_turn: bool,
+    compaction_count: u64,
+}
+
+fn finalize_window_generation_segment(
+    active_segment: WindowGenerationReplaySegment,
+    window_generation: &mut u64,
+    pending_rollback_turns: &mut usize,
+) {
+    if *pending_rollback_turns > 0 {
+        if active_segment.counts_as_user_turn {
+            *pending_rollback_turns -= 1;
+        }
+        return;
+    }
+
+    *window_generation = window_generation.saturating_add(active_segment.compaction_count);
+}
+
+/// Replays rollout segments newest-to-oldest so compactions in rolled-back suffixes do not
+/// contribute to the public context-window lineage generation.
+pub(super) fn effective_window_generation_from_rollout(rollout_items: &[RolloutItem]) -> u64 {
+    let mut window_generation = 0u64;
+    let mut pending_rollback_turns = 0usize;
+    let mut active_segment: Option<WindowGenerationReplaySegment> = None;
+
+    for item in rollout_items.iter().rev() {
+        match item {
+            RolloutItem::Compacted(_) => {
+                let active_segment =
+                    active_segment.get_or_insert_with(WindowGenerationReplaySegment::default);
+                active_segment.compaction_count = active_segment.compaction_count.saturating_add(1);
+            }
+            RolloutItem::EventMsg(EventMsg::ThreadRolledBack(rollback)) => {
+                pending_rollback_turns = pending_rollback_turns
+                    .saturating_add(usize::try_from(rollback.num_turns).unwrap_or(usize::MAX));
+            }
+            RolloutItem::EventMsg(EventMsg::TurnStarted(_)) => {
+                if let Some(active_segment) = active_segment.take() {
+                    finalize_window_generation_segment(
+                        active_segment,
+                        &mut window_generation,
+                        &mut pending_rollback_turns,
+                    );
+                }
+            }
+            RolloutItem::EventMsg(EventMsg::UserMessage(_)) => {
+                active_segment
+                    .get_or_insert_with(WindowGenerationReplaySegment::default)
+                    .counts_as_user_turn = true;
+            }
+            RolloutItem::ResponseItem(response_item) => {
+                let active_segment =
+                    active_segment.get_or_insert_with(WindowGenerationReplaySegment::default);
+                active_segment.counts_as_user_turn |= is_user_turn_boundary(response_item);
+            }
+            RolloutItem::EventMsg(_)
+            | RolloutItem::TurnContext(_)
+            | RolloutItem::SessionMeta(_) => {}
+        }
+    }
+
+    if let Some(active_segment) = active_segment {
+        finalize_window_generation_segment(
+            active_segment,
+            &mut window_generation,
+            &mut pending_rollback_turns,
+        );
+    }
+
+    window_generation
+}
+
 fn turn_ids_are_compatible(active_turn_id: Option<&str>, item_turn_id: Option<&str>) -> bool {
     active_turn_id
         .is_none_or(|turn_id| item_turn_id.is_none_or(|item_turn_id| item_turn_id == turn_id))
diff --git a/codex-rs/core/src/session/session.rs b/codex-rs/core/src/session/session.rs
index 37a501fb38..6915665663 100644
--- a/codex-rs/core/src/session/session.rs
+++ b/codex-rs/core/src/session/session.rs
@@ -518,15 +518,15 @@ impl Session {
             InitialHistory::Resumed(resumed_history) => resumed_history.conversation_id,
         };
         let window_generation = match &initial_history {
-            InitialHistory::Resumed(resumed_history) => u64::try_from(
-                resumed_history
-                    .history
-                    .iter()
-                    .filter(|item| matches!(item, RolloutItem::Compacted(_)))
-                    .count(),
-            )
-            .unwrap_or(u64::MAX),
-            InitialHistory::New | InitialHistory::Cleared | InitialHistory::Forked(_) => 0,
+            InitialHistory::Resumed(resumed_history) => {
+                super::rollout_reconstruction::effective_window_generation_from_rollout(
+                    &resumed_history.history,
+                )
+            }
+            InitialHistory::Forked(history) => {
+                super::rollout_reconstruction::effective_window_generation_from_rollout(history)
+            }
+            InitialHistory::New | InitialHistory::Cleared => 0,
         };
         // Kick off independent async setup tasks in parallel to reduce startup latency.
         //
diff --git a/codex-rs/core/tests/suite/window_headers.rs b/codex-rs/core/tests/suite/window_headers.rs
index 51bb132799..5115a007a8 100644
--- a/codex-rs/core/tests/suite/window_headers.rs
+++ b/codex-rs/core/tests/suite/window_headers.rs
@@ -164,6 +164,162 @@ async fn window_id_stays_stable_after_rollback_and_resume() -> Result<()> {
     Ok(())
 }
 
+#[tokio::test(flavor = "multi_thread", worker_threads = 2)]
+async fn window_id_rolls_back_across_compaction_and_persists_on_resume() -> Result<()> {
+    skip_if_no_network!(Ok(()));
+
+    let server = start_mock_server().await;
+    let request_log = mount_sse_sequence(
+        &server,
+        vec![
+            sse(vec![
+                ev_assistant_message("msg-1", "first reply"),
+                ev_completed("resp-1"),
+            ]),
+            sse(vec![
+                ev_assistant_message("msg-2", "summary"),
+                ev_completed("resp-2"),
+            ]),
+            sse(vec![ev_completed("resp-3")]),
+            sse(vec![ev_completed("resp-4")]),
+            sse(vec![ev_completed("resp-5")]),
+        ],
+    )
+    .await;
+
+    let mut builder = test_codex().with_config(|config| {
+        config.model_provider.name = "Non-OpenAI Model provider".to_string();
+        config.compact_prompt = Some(SUMMARIZATION_PROMPT.to_string());
+    });
+    let initial = builder.build(&server).await?;
+    let initial_thread = Arc::clone(&initial.codex);
+    let rollout_path = initial
+        .session_configured
+        .rollout_path
+        .clone()
+        .expect("rollout path");
+
+    submit_user_turn(&initial_thread, "before compact").await?;
+    submit_compact_turn(&initial_thread).await?;
+    submit_user_turn(&initial_thread, "discard me").await?;
+    initial_thread
+        .submit(Op::ThreadRollback { num_turns: 2 })
+        .await?;
+    wait_for_event(&initial_thread, |event| {
+        matches!(event, EventMsg::ThreadRolledBack(_))
+    })
+    .await;
+    submit_user_turn(&initial_thread, "after rollback").await?;
+    shutdown_thread(&initial_thread).await?;
+
+    let resumed = builder
+        .resume(&server, initial.home.clone(), rollout_path)
+        .await?;
+    submit_user_turn(&resumed.codex, "after resume").await?;
+    shutdown_thread(&resumed.codex).await?;
+
+    let requests = request_log.requests();
+    assert_eq!(requests.len(), 5, "expected five model requests");
+
+    let window_ids = requests.iter().map(window_id_parts).collect::<Vec<_>>();
+    let initial_thread_id = window_ids[0].0.clone();
+    assert_eq!(
+        window_ids,
+        vec![
+            (initial_thread_id.clone(), 0),
+            (initial_thread_id.clone(), 0),
+            (initial_thread_id.clone(), 1),
+            (initial_thread_id.clone(), 0),
+            (initial_thread_id, 0),
+        ]
+    );
+
+    Ok(())
+}
+
+#[tokio::test(flavor = "multi_thread", worker_threads = 2)]
+async fn forked_compacted_history_inherits_effective_generation_on_resume() -> Result<()> {
+    skip_if_no_network!(Ok(()));
+
+    let server = start_mock_server().await;
+    let request_log = mount_sse_sequence(
+        &server,
+        vec![
+            sse(vec![
+                ev_assistant_message("msg-1", "first reply"),
+                ev_completed("resp-1"),
+            ]),
+            sse(vec![
+                ev_assistant_message("msg-2", "summary"),
+                ev_completed("resp-2"),
+            ]),
+            sse(vec![ev_completed("resp-3")]),
+            sse(vec![ev_completed("resp-4")]),
+        ],
+    )
+    .await;
+
+    let mut builder = test_codex().with_config(|config| {
+        config.model_provider.name = "Non-OpenAI Model provider".to_string();
+        config.compact_prompt = Some(SUMMARIZATION_PROMPT.to_string());
+    });
+    let initial = builder.build(&server).await?;
+    let initial_thread = Arc::clone(&initial.codex);
+    let rollout_path = initial
+        .session_configured
+        .rollout_path
+        .clone()
+        .expect("rollout path");
+
+    submit_user_turn(&initial_thread, "before compact").await?;
+    submit_compact_turn(&initial_thread).await?;
+    shutdown_thread(&initial_thread).await?;
+
+    let forked = initial
+        .thread_manager
+        .fork_thread(
+            /*snapshot*/ usize::MAX,
+            initial.config.clone(),
+            rollout_path,
+            /*thread_source*/ None,
+            /*persist_extended_history*/ false,
+            /*parent_trace*/ None,
+        )
+        .await?;
+    let fork_rollout_path = forked
+        .session_configured
+        .rollout_path
+        .clone()
+        .expect("fork rollout path");
+    submit_user_turn(&forked.thread, "after fork").await?;
+    shutdown_thread(&forked.thread).await?;
+
+    let resumed = builder
+        .resume(&server, initial.home.clone(), fork_rollout_path)
+        .await?;
+    submit_user_turn(&resumed.codex, "after fork resume").await?;
+    shutdown_thread(&resumed.codex).await?;
+
+    let requests = request_log.requests();
+    assert_eq!(requests.len(), 4, "expected four model requests");
+
+    let window_ids = requests.iter().map(window_id_parts).collect::<Vec<_>>();
+    let initial_thread_id = window_ids[0].0.clone();
+    let forked_thread_id = window_ids[2].0.clone();
+    assert_ne!(forked_thread_id, initial_thread_id);
+    assert_eq!(
+        window_ids,
+        vec![
+            (initial_thread_id.clone(), 0),
+            (initial_thread_id, 0),
+            (forked_thread_id.clone(), 1),
+            (forked_thread_id, 1),
+        ]
+    );
+
+    Ok(())
+}
+
 async fn submit_user_turn(codex: &Arc<CodexThread>, text: &str) -> Result<()> {
     codex
         .submit(Op::UserInput {