diff --git a/src/ex/thread_pool.cpp b/src/ex/thread_pool.cpp
index a5a0c1e76..e361c183c 100644
--- a/src/ex/thread_pool.cpp
+++ b/src/ex/thread_pool.cpp
@@ -110,6 +110,15 @@ class thread_pool::impl
         cv_.notify_one();
     }
 
+    void
+    join() noexcept
+    {
+        stop();
+        for(auto& t : threads_)
+            if(t.joinable())
+                t.join();
+    }
+
     void
     stop() noexcept
     {
@@ -160,6 +169,7 @@ class thread_pool::impl
 thread_pool::
 ~thread_pool()
 {
+    impl_->join();
     shutdown();
     destroy();
     delete impl_;
diff --git a/test/unit/when_all.cpp b/test/unit/when_all.cpp
index a7b22614e..0ad92599c 100644
--- a/test/unit/when_all.cpp
+++ b/test/unit/when_all.cpp
@@ -12,12 +12,15 @@
 
 #include <boost/capy/ex/async_event.hpp>
 #include <boost/capy/ex/run_async.hpp>
+#include <boost/capy/ex/strand.hpp>
+#include <boost/capy/ex/thread_pool.hpp>
 #include <boost/capy/io_task.hpp>
 #include <boost/capy/task.hpp>
 
 #include "test_helpers.hpp"
 
 #include <atomic>
+#include <latch>
 #include <stdexcept>
 #include <string>
 #include <vector>
@@ -1047,5 +1050,85 @@ TEST_SUITE(
     when_all_io_awaitable_test,
     "boost.capy.when_all_io_awaitable");
 
+struct when_all_strand_test
+{
+    // Regression for #131: executor_ref::dispatch() formerly
+    // returned void, discarding the symmetric transfer handle
+    // from strand::dispatch(). This caused when_all child
+    // runners to never resume, deadlocking the caller.
+    void
+    testStrandWhenAll()
+    {
+        thread_pool pool(2);
+        strand s{pool.get_executor()};
+        std::latch done(1);
+        bool completed = false;
+
+        auto outer = [&]() -> task<> {
+            co_await when_all(
+                []() -> task<> { co_return; }(),
+                []() -> task<> { co_return; }()
+            );
+        };
+
+        run_async(s,
+            [&](auto&&...) {
+                completed = true;
+                done.count_down();
+            },
+            [&](auto) {
+                done.count_down();
+            }
+        )(outer());
+
+        done.wait();
+        BOOST_TEST(completed);
+    }
+
+    // Verify strand + when_all propagates values correctly
+    void
+    testStrandWhenAllWithValues()
+    {
+        thread_pool pool(2);
+        strand s{pool.get_executor()};
+        std::latch done(1);
+        bool completed = false;
+        int result = 0;
+
+        auto outer = [&]() -> task<std::tuple<int, int>> {
+            co_return co_await when_all(
+                returns_int(10),
+                returns_int(20));
+        };
+
+        run_async(s,
+            [&](std::tuple<int, int> t) {
+                auto [a, b] = t;
+                completed = true;
+                result = a + b;
+                done.count_down();
+            },
+            [&](auto) {
+                done.count_down();
+            }
+        )(outer());
+
+        done.wait();
+        BOOST_TEST(completed);
+        BOOST_TEST_EQ(result, 30);
+    }
+
+    void
+    run()
+    {
+        testStrandWhenAll();
+        testStrandWhenAllWithValues();
+    }
+};
+
+TEST_SUITE(
+    when_all_strand_test,
+    "boost.capy.when_all_strand");
+
 } // capy
 } // boost