diff --git a/clients/ios/App/IOSThreadStore.swift b/clients/ios/App/IOSThreadStore.swift
index 2f25add1cf3..9f27b904469 100644
--- a/clients/ios/App/IOSThreadStore.swift
+++ b/clients/ios/App/IOSThreadStore.swift
@@ -470,7 +470,15 @@ class IOSThreadStore: ObservableObject {
     /// Request an older page of history for pagination.
     private func requestPaginatedHistory(sessionId: String, beforeTimestamp: Double) {
         guard let daemon = daemonClient as? DaemonClient,
-              let thread = threads.first(where: { $0.sessionId == sessionId }) else { return }
+              let thread = threads.first(where: { $0.sessionId == sessionId }) else {
+            // Clear loading state so the user isn't stuck with a permanent spinner.
+            // The daemon cast may fail (e.g. HTTP transport) while the thread is still findable.
+            if let thread = threads.first(where: { $0.sessionId == sessionId }),
+               let vm = viewModels[thread.id] {
+                vm.isLoadingMoreMessages = false
+            }
+            return
+        }
         pendingHistoryBySessionId[sessionId] = thread.id
         do {
             try daemon.sendHistoryRequest(sessionId: sessionId, limit: 50, beforeTimestamp: beforeTimestamp, mode: "light")
diff --git a/clients/macos/vellum-assistant/Features/MainWindow/ThreadManager.swift b/clients/macos/vellum-assistant/Features/MainWindow/ThreadManager.swift
index 02413149d92..83482e331f0 100644
--- a/clients/macos/vellum-assistant/Features/MainWindow/ThreadManager.swift
+++ b/clients/macos/vellum-assistant/Features/MainWindow/ThreadManager.swift
@@ -513,6 +513,14 @@ final class ThreadManager: ObservableObject, ThreadRestorerDelegate {
         return vm
     }
 
+    func existingChatViewModel(forSessionId sessionId: String) -> ChatViewModel? {
+        for (threadId, vm) in chatViewModels where vm.sessionId == sessionId {
+            touchVMAccessOrder(threadId)
+            return vm
+        }
+        return nil
+    }
+
     func setChatViewModel(_ vm: ChatViewModel, for threadId: UUID) {
         chatViewModels[threadId] = vm
         touchVMAccessOrder(threadId)
diff --git a/clients/macos/vellum-assistant/Features/MainWindow/ThreadSessionRestorer.swift b/clients/macos/vellum-assistant/Features/MainWindow/ThreadSessionRestorer.swift
index 35b126c49f7..8ea9185cb76 100644
--- a/clients/macos/vellum-assistant/Features/MainWindow/ThreadSessionRestorer.swift
+++ b/clients/macos/vellum-assistant/Features/MainWindow/ThreadSessionRestorer.swift
@@ -26,6 +26,8 @@ protocol ThreadRestorerDelegate: AnyObject {
     func isSessionArchived(_ sessionId: String) -> Bool
     func restoreLastActiveThread()
     func appendThreads(from response: SessionListResponseMessage)
+    /// Returns an existing ChatViewModel matching the given session ID, if any.
+    func existingChatViewModel(forSessionId sessionId: String) -> ChatViewModel?
 }
 
 /// Handles daemon session restoration: fetching the session list on connect,
@@ -126,7 +128,12 @@ final class ThreadSessionRestorer {
     /// trigger in the message list when all locally loaded messages are visible.
     func requestPaginatedHistory(sessionId: String, beforeTimestamp: Double) {
         guard let delegate else { return }
-        guard let thread = delegate.threads.first(where: { $0.sessionId == sessionId }) else { return }
+        guard let thread = delegate.threads.first(where: { $0.sessionId == sessionId }) else {
+            // Thread removed from the list during a concurrent reconnect/refresh.
+            // Reset loading state so the user isn't stuck with a permanent spinner.
+            delegate.existingChatViewModel(forSessionId: sessionId)?.isLoadingMoreMessages = false
+            return
+        }
         pendingHistoryBySessionId[sessionId] = thread.id
         do {
             try daemonClient.sendHistoryRequest(sessionId: sessionId, limit: 50, beforeTimestamp: beforeTimestamp, mode: "light")
diff --git a/clients/macos/vellum-assistantTests/ThreadSessionRestorerTests.swift b/clients/macos/vellum-assistantTests/ThreadSessionRestorerTests.swift
index 3676fe1a657..135f712270c 100644
--- a/clients/macos/vellum-assistantTests/ThreadSessionRestorerTests.swift
+++ b/clients/macos/vellum-assistantTests/ThreadSessionRestorerTests.swift
@@ -30,6 +30,13 @@ final class MockThreadRestorerDelegate: ThreadRestorerDelegate {
         viewModels[threadId]
     }
 
+    func existingChatViewModel(forSessionId sessionId: String) -> ChatViewModel? {
+        for (_, vm) in viewModels where vm.sessionId == sessionId {
+            return vm
+        }
+        return nil
+    }
+
     func setChatViewModel(_ vm: ChatViewModel, for threadId: UUID) {
         viewModels[threadId] = vm
     }
diff --git a/clients/shared/Features/Chat/ChatViewModel.swift b/clients/shared/Features/Chat/ChatViewModel.swift
index 9ac6f2b6f23..7dc380407d8 100644
--- a/clients/shared/Features/Chat/ChatViewModel.swift
+++ b/clients/shared/Features/Chat/ChatViewModel.swift
@@ -390,6 +390,9 @@ public final class ChatViewModel: ObservableObject {
     /// True while a previous-page load is in progress (brief async delay for UX).
     @Published public var isLoadingMoreMessages: Bool = false
 
+    /// Timeout task that resets `isLoadingMoreMessages` if the daemon never responds.
+    private var loadMoreTimeoutTask: Task<Void, Never>?
+
     /// The subset of messages that are actually displayed (excludes subagent notifications
     /// and other UI-only messages that the view filters before rendering).
     public var displayedMessages: [ChatMessage] { messages.filter { !$0.isSubagentNotification } }
@@ -446,6 +449,14 @@ public final class ChatViewModel: ObservableObject {
         // All local messages are visible — fetch the next page from the daemon.
         guard hasMoreHistory, let cursor = historyCursor, let sessionId else { return false }
         isLoadingMoreMessages = true
+        // Safety timeout: if the daemon drops the connection or never responds,
+        // reset the flag so the user can retry.
+        loadMoreTimeoutTask?.cancel()
+        loadMoreTimeoutTask = Task { @MainActor [weak self] in
+            try? await Task.sleep(nanoseconds: 15_000_000_000) // 15 seconds
+            guard let self, self.isLoadingMoreMessages else { return }
+            self.isLoadingMoreMessages = false
+        }
         onLoadMoreHistory?(sessionId, cursor)
         // The loading indicator is cleared by populateFromHistory when the response arrives.
         return true
@@ -1936,6 +1947,8 @@ public final class ChatViewModel: ObservableObject {
             } else {
                 displayedMessageCount = Int.max
             }
+            self.loadMoreTimeoutTask?.cancel()
+            self.loadMoreTimeoutTask = nil
             self.isLoadingMoreMessages = false
             trimOldMessagesIfNeeded()
             return