diff --git a/Kernel/src/renderive/scene/base/Scene_Base.cpp b/Kernel/src/renderive/scene/base/Scene_Base.cpp
index 2b77ae4..f8eb0b3 100644
--- a/Kernel/src/renderive/scene/base/Scene_Base.cpp
+++ b/Kernel/src/renderive/scene/base/Scene_Base.cpp
@@ -1,5 +1,6 @@
 #include "Scene_Base.hpp"
 #include <algorithm>
+#include <chrono>
 #include <memory_resource>
 #include <stdexcept>
 #include <string>
@@ -11,6 +12,12 @@
 #include "renderive/state/base/State_Strategy_Base.hpp"
 static_assert(TF_VERSION == 400100, "Renderive requires Taskflow 4.1.0");
 namespace {
+std::uint64_t steady_now_ns() noexcept {
+    return static_cast<std::uint64_t>(
+        std::chrono::duration_cast<std::chrono::nanoseconds>(
+            std::chrono::steady_clock::now().time_since_epoch()).count());
+}
+
 template <class Node_Accessor>
 Scene_Base::Renderable_List topological_order(
     const Scene_Base::Renderable_List& renderables,
@@ -331,6 +338,50 @@ std::vector<Scene_Base::Const_Renderable> Scene_Base::paint_order_snapshot() con
     const auto ordered = display_order(*cache_renderables_);
     return {ordered.begin(), ordered.end()};
 }
+Task_Execution_Snapshot Scene_Base::task_execution_snapshot() const noexcept {
+    Task_Execution_Snapshot snapshot;
+    snapshot.worker_count = task_executor_worker_count();
+    snapshot.render_sequence =
+        task_execution_render_sequence_.load(std::memory_order_acquire);
+    snapshot.graph_node_count =
+        task_execution_node_count_.load(std::memory_order_acquire);
+    snapshot.completed_count =
+        task_execution_completed_count_.load(std::memory_order_acquire);
+    snapshot.running_count =
+        task_execution_running_count_.load(std::memory_order_acquire);
+    snapshot.peak_parallelism =
+        task_execution_peak_parallelism_.load(std::memory_order_acquire);
+    snapshot.failed_count =
+        task_execution_failed_count_.load(std::memory_order_acquire);
+    const std::size_t accounted = snapshot.completed_count + snapshot.running_count;
+    snapshot.pending_count = accounted < snapshot.graph_node_count
+                                 ? snapshot.graph_node_count - accounted
+                                 : 0;
+    {
+        std::lock_guard lock(task_mutex_);
+        if (task_pending_)
+            snapshot.state = Task_Execution_State::Queued;
+        else if (rendering_)
+            snapshot.state = Task_Execution_State::Running;
+        else if (current_completion_ && current_completion_->completed)
+            snapshot.state = current_completion_->exception
+                                 ? Task_Execution_State::Failed
+                                 : Task_Execution_State::Completed;
+    }
+    snapshot.duration_ns =
+        task_execution_duration_ns_.load(std::memory_order_acquire);
+    if (snapshot.state == Task_Execution_State::Running) {
+        const auto started =
+            task_execution_started_ns_.load(std::memory_order_acquire);
+        const auto now = steady_now_ns();
+        if (started != 0 && now >= started)
+            snapshot.duration_ns = now - started;
+    }
+    return snapshot;
+}
+std::size_t Scene_Base::task_executor_worker_count() noexcept {
+    return Execution_Context::executor().num_workers();
+}
 std::pmr::memory_resource& Scene_Base::memory_resource() const noexcept {
     return memory_domain_->resource();
 }
@@ -459,6 +510,20 @@ void Scene_Base::execute_taskflow(Render_Task& task) {
     };
     std::pmr::monotonic_buffer_resource scratch_resource(&memory_resource());
     tf::Taskflow scene_graph;
+    std::size_t planned_task_count{};
+    const auto tracked = [this, &planned_task_count](auto function) {
+        ++planned_task_count;
+        return [this, function = std::move(function)]() mutable {
+            task_execution_started();
+            try {
+                function();
+                task_execution_finished(false);
+            } catch (...) {
+                task_execution_finished(true);
+                throw;
+            }
+        };
+    };
     std::pmr::unordered_map<Renderable_Base*, std::size_t> indices(&scratch_resource);
     indices.reserve(task.render_order.size());
     std::pmr::vector<bool> render_flags(task.render_order.size(), false, &scratch_resource);
@@ -500,21 +565,21 @@ void Scene_Base::execute_taskflow(Render_Task& task) {
             std::pmr::vector<tf::Task> internal_tasks(&scratch_resource);
             internal_tasks.reserve(nodes.size());
             for (const auto& node : nodes) {
-                internal_tasks.push_back(module.graph.emplace([this, function = node.function, context] {
+                internal_tasks.push_back(module.graph.emplace(tracked([this, function = node.function, context] {
                     Render_Execution_Scope scope(*this);
                     function(context);
-                }).name(std::string(node.name)));
+                })).name(std::string(node.name)));
             }
             for (std::size_t node_index = 0; node_index < nodes.size(); ++node_index) {
                 for (std::size_t successor : nodes[node_index].successors) {
                     internal_tasks[node_index].precede(internal_tasks.at(successor));
                 }
             }
-            tf::Task mark_rendered = module.graph.emplace([&renderable, render_revision] {
+            tf::Task mark_rendered = module.graph.emplace(tracked([&renderable, render_revision] {
                 renderable.mark_rendered(render_revision);
-            }).name("cache_publish");
+            })).name("cache_publish");
             if (internal_tasks.empty()) {
-                module.graph.emplace([] {}).precede(mark_rendered);
+                module.graph.emplace(tracked([] {})).precede(mark_rendered);
             } else {
                 for (std::size_t node_index = 0; node_index < nodes.size(); ++node_index) {
                     if (nodes[node_index].successors.empty()) {
@@ -523,7 +588,7 @@ void Scene_Base::execute_taskflow(Render_Task& task) {
                 }
             }
         } else {
-            module.graph.emplace([] {}).name("cache_hit");
+            module.graph.emplace(tracked([] {})).name("cache_hit");
         }
         module.module_task = scene_graph.composed_of(module.graph).name("renderable");
     }
@@ -538,18 +603,62 @@ void Scene_Base::execute_taskflow(Render_Task& task) {
         }
     }
     const Scene_Render_Context final_context{this, task.frame_control_state, task.scene_state_revision, task.render_sequence, nullptr, nullptr};
-    tf::Task final_task = scene_graph.emplace([this, final_context, &task] {
+    tf::Task final_task = scene_graph.emplace(tracked([this, final_context, &task] {
         Render_Execution_Scope scope(*this);
         generate_final_color_cache(final_context, task.display_order);
-    }).name("final_color_cache");
+    })).name("final_color_cache");
     if (modules.empty()) {
-        scene_graph.emplace([] {}).precede(final_task);
+        scene_graph.emplace(tracked([] {})).precede(final_task);
     } else {
         for (Module& module : modules) {
             module.module_task.precede(final_task);
         }
     }
-    Execution_Context::executor().run(scene_graph).get();
+    begin_task_execution(task.render_sequence);
+    set_task_execution_node_count(planned_task_count);
+    try {
+        Execution_Context::executor().run(scene_graph).get();
+    } catch (...) {
+        finish_task_execution();
+        throw;
+    }
+    finish_task_execution();
+}
+void Scene_Base::begin_task_execution(std::uint64_t render_sequence) noexcept {
+    task_execution_render_sequence_.store(render_sequence, std::memory_order_release);
+    task_execution_node_count_.store(0, std::memory_order_release);
+    task_execution_completed_count_.store(0, std::memory_order_release);
+    task_execution_running_count_.store(0, std::memory_order_release);
+    task_execution_peak_parallelism_.store(0, std::memory_order_release);
+    task_execution_failed_count_.store(0, std::memory_order_release);
+    task_execution_duration_ns_.store(0, std::memory_order_release);
+    task_execution_started_ns_.store(steady_now_ns(), std::memory_order_release);
+}
+void Scene_Base::set_task_execution_node_count(std::size_t count) noexcept {
+    task_execution_node_count_.store(count, std::memory_order_release);
+}
+void Scene_Base::task_execution_started() noexcept {
+    const std::size_t active =
+        task_execution_running_count_.fetch_add(1, std::memory_order_acq_rel) + 1;
+    std::size_t peak =
+        task_execution_peak_parallelism_.load(std::memory_order_acquire);
+    while (peak < active &&
+           !task_execution_peak_parallelism_.compare_exchange_weak(
+               peak, active, std::memory_order_acq_rel, std::memory_order_acquire)) {}
+}
+void Scene_Base::task_execution_finished(bool failed) noexcept {
+    if (failed)
+        task_execution_failed_count_.fetch_add(1, std::memory_order_acq_rel);
+    task_execution_completed_count_.fetch_add(1, std::memory_order_acq_rel);
+    task_execution_running_count_.fetch_sub(1, std::memory_order_acq_rel);
+}
+void Scene_Base::finish_task_execution() noexcept {
+    const auto started =
+        task_execution_started_ns_.load(std::memory_order_acquire);
+    const auto now = steady_now_ns();
+    task_execution_duration_ns_.store(
+        started != 0 && now >= started ? now - started : 0,
+        std::memory_order_release);
 }
 void Scene_Base::validate_renderable_scene(const Renderable_Base& renderable) const {
     if (renderable.real_time_data_state_->scene_lifetime.get() != scene_lifetime_.get()) {
diff --git a/Kernel/src/renderive/scene/base/Scene_Base.hpp b/Kernel/src/renderive/scene/base/Scene_Base.hpp
index ff975dd..22fb814 100644
--- a/Kernel/src/renderive/scene/base/Scene_Base.hpp
+++ b/Kernel/src/renderive/scene/base/Scene_Base.hpp
@@ -1,5 +1,6 @@
 #pragma once
 #include <array>
+#include <atomic>
 #include <condition_variable>
 #include <concepts>
 #include <cstdint>
@@ -15,6 +16,7 @@
 #include "renderive/renderable/base/Renderable_Base.hpp"
 #include "Scene_Lifetime.hpp"
 #include "Scene_Render_Context.hpp"
+#include "Task_Execution_Snapshot.hpp"
 class Color_Cache;
 class Scene_Base {
 private:
@@ -65,6 +67,8 @@ public:
     std::size_t renderable_count() const;
     Topology_Snapshot topology_snapshot() const;
     std::vector<Const_Renderable> paint_order_snapshot() const;
+    [[nodiscard]] Task_Execution_Snapshot task_execution_snapshot() const noexcept;
+    [[nodiscard]] static std::size_t task_executor_worker_count() noexcept;
     std::pmr::memory_resource& memory_resource() const noexcept;
     std::pmr::memory_resource& upstream_memory_resource() const noexcept;
     Frame_Control_Strategy_Base& frame_control_strategy();
@@ -114,6 +118,11 @@ private:
     };
     void render_loop();
     void execute_taskflow(Render_Task& task);
+    void begin_task_execution(std::uint64_t render_sequence) noexcept;
+    void set_task_execution_node_count(std::size_t count) noexcept;
+    void task_execution_started() noexcept;
+    void task_execution_finished(bool failed) noexcept;
+    void finish_task_execution() noexcept;
     void validate_renderable_scene(const Renderable_Base& renderable) const;
     bool is_renderable_attached_locked(const Renderable_Base& renderable) const;
     void validate_renderable_attached_locked(const Renderable_Base& renderable) const;
@@ -125,7 +134,7 @@ private:
     Renderable_List* render_renderables_;
     Renderable_List* cache_renderables_;
     mutable std::mutex renderable_mutex_;
-    std::recursive_mutex task_mutex_;
+    mutable std::recursive_mutex task_mutex_;
     std::condition_variable_any task_ready_;
     std::condition_variable_any render_completed_;
     std::thread worker_;
@@ -138,6 +147,14 @@ private:
     bool task_pending_{};
     bool rendering_{};
     bool stop_{};
+    std::atomic<std::uint64_t> task_execution_render_sequence_{};
+    std::atomic<std::size_t> task_execution_node_count_{};
+    std::atomic<std::size_t> task_execution_completed_count_{};
+    std::atomic<std::size_t> task_execution_running_count_{};
+    std::atomic<std::size_t> task_execution_peak_parallelism_{};
+    std::atomic<std::size_t> task_execution_failed_count_{};
+    std::atomic<std::uint64_t> task_execution_started_ns_{};
+    std::atomic<std::uint64_t> task_execution_duration_ns_{};
 };
 class Scene_2D_Base : public Scene_Base {
 public:
diff --git a/Kernel/src/renderive/scene/base/Task_Execution_Snapshot.hpp b/Kernel/src/renderive/scene/base/Task_Execution_Snapshot.hpp
new file mode 100644
index 0000000..337de0e
--- /dev/null
+++ b/Kernel/src/renderive/scene/base/Task_Execution_Snapshot.hpp
@@ -0,0 +1,25 @@
+#pragma once
+
+#include <cstddef>
+#include <cstdint>
+
+enum class Task_Execution_State : std::uint8_t {
+    Idle,
+    Queued,
+    Running,
+    Completed,
+    Failed
+};
+
+struct Task_Execution_Snapshot {
+    Task_Execution_State state{Task_Execution_State::Idle};
+    std::size_t worker_count{};
+    std::uint64_t render_sequence{};
+    std::size_t graph_node_count{};
+    std::size_t completed_count{};
+    std::size_t running_count{};
+    std::size_t pending_count{};
+    std::size_t peak_parallelism{};
+    std::size_t failed_count{};
+    std::uint64_t duration_ns{};
+};
diff --git a/Kernel/tests/renderive/renderable/Renderable_Task_Graph_Test.cpp b/Kernel/tests/renderive/renderable/Renderable_Task_Graph_Test.cpp
index 93c3af9..74ae025 100644
--- a/Kernel/tests/renderive/renderable/Renderable_Task_Graph_Test.cpp
+++ b/Kernel/tests/renderive/renderable/Renderable_Task_Graph_Test.cpp
@@ -129,8 +129,25 @@ TEST(renderable_task_graph_concurrency_test, runs_independent_internal_tasks_con
     auto renderable = std::make_shared<Renderable_Task_Graph_Concurrency_Test_Renderable>(scene, active, maximum);
     scene.attach_renderable(renderable);
     scene.render();
+    while (active.load(std::memory_order_acquire) == 0)
+        std::this_thread::yield();
+    const auto running = scene.task_execution_snapshot();
+    EXPECT_EQ(running.state, Task_Execution_State::Running);
+    EXPECT_EQ(running.graph_node_count, 4u);
+    EXPECT_GT(running.running_count, 0u);
+    EXPECT_LT(running.completed_count, running.graph_node_count);
     scene.wait_for_render();
     EXPECT_GE(maximum.load(std::memory_order_acquire), 2);
+    const auto execution = scene.task_execution_snapshot();
+    EXPECT_EQ(execution.state, Task_Execution_State::Completed);
+    EXPECT_EQ(execution.worker_count, Scene_Base::task_executor_worker_count());
+    EXPECT_EQ(execution.graph_node_count, 4u);
+    EXPECT_EQ(execution.completed_count, execution.graph_node_count);
+    EXPECT_EQ(execution.running_count, 0u);
+    EXPECT_EQ(execution.pending_count, 0u);
+    EXPECT_GE(execution.peak_parallelism, 2u);
+    EXPECT_EQ(execution.failed_count, 0u);
+    EXPECT_GT(execution.duration_ns, 0u);
 }
 #endif
 TEST(renderable_task_graph_test, rejects_tasks_from_different_graphs) {
diff --git a/Kernel/tests/renderive/scene/base/Scene_Base_Test.cpp b/Kernel/tests/renderive/scene/base/Scene_Base_Test.cpp
index 55b6e4b..1ce7c19 100644
--- a/Kernel/tests/renderive/scene/base/Scene_Base_Test.cpp
+++ b/Kernel/tests/renderive/scene/base/Scene_Base_Test.cpp
@@ -38,6 +38,12 @@ TEST(scene_base_test, wait_for_render_propagates_background_render_failure) {
     scene.attach_renderable(renderable);
     scene.render();
     EXPECT_THROW(scene.wait_for_render(), std::runtime_error);
+    const auto execution = scene.task_execution_snapshot();
+    EXPECT_EQ(execution.state, Task_Execution_State::Failed);
+    EXPECT_EQ(execution.graph_node_count, 3u);
+    EXPECT_EQ(execution.failed_count, 1u);
+    EXPECT_EQ(execution.running_count, 0u);
+    EXPECT_GT(execution.duration_ns, 0u);
 }
 TEST(scene_base_test, scene_destruction_does_not_throw_after_background_render_failure) {
     {
diff --git a/render_2D/plot/Plot_Core.cpp b/render_2D/plot/Plot_Core.cpp
index 5007707..d970abc 100644
--- a/render_2D/plot/Plot_Core.cpp
+++ b/render_2D/plot/Plot_Core.cpp
@@ -553,6 +553,18 @@ Frame_Observer_Snapshot Plot_Core::frame_observer_snapshot() const {
         snapshot.next_refresh_interval_ns = state.next_refresh_interval_ns;
         snapshot.end_to_end_ns = state.end_to_end_ns;
     }
+    const auto task_execution = with_scene(
+        impl_->scene,
+        [](const auto& scene) { return scene.task_execution_snapshot(); });
+    snapshot.task_execution_state = task_execution.state;
+    snapshot.task_worker_count = task_execution.worker_count;
+    snapshot.task_graph_node_count = task_execution.graph_node_count;
+    snapshot.task_completed_node_count = task_execution.completed_count;
+    snapshot.task_running_node_count = task_execution.running_count;
+    snapshot.task_pending_node_count = task_execution.pending_count;
+    snapshot.task_peak_parallelism = task_execution.peak_parallelism;
+    snapshot.task_failed_node_count = task_execution.failed_count;
+    snapshot.task_graph_duration_ns = task_execution.duration_ns;
     return snapshot;
 }
 
diff --git a/render_2D/plot/Plot_Core.h b/render_2D/plot/Plot_Core.h
index c10a5b4..392699a 100644
--- a/render_2D/plot/Plot_Core.h
+++ b/render_2D/plot/Plot_Core.h
@@ -3,6 +3,7 @@
 #include "../base/Types.h"
 #include "../event/Event.h"
 #include <renderive/frame_control/Frame_Consumer_Feedback.hpp>
+#include <renderive/scene/base/Task_Execution_Snapshot.hpp>
 
 #include <atomic>
 #include <concepts>
@@ -69,6 +70,15 @@ struct Frame_Observer_Snapshot {
     std::uint64_t render_finish_state_wait_ns{};
     std::uint64_t queue_wait_ns{};
     std::uint64_t end_to_end_ns{};
+    Task_Execution_State task_execution_state{Task_Execution_State::Idle};
+    std::size_t task_worker_count{};
+    std::size_t task_graph_node_count{};
+    std::size_t task_completed_node_count{};
+    std::size_t task_running_node_count{};
+    std::size_t task_pending_node_count{};
+    std::size_t task_peak_parallelism{};
+    std::size_t task_failed_node_count{};
+    std::uint64_t task_graph_duration_ns{};
 };
 
 class Presentation_Sink {
diff --git a/render_2D/plottable/Afterglow.cpp b/render_2D/plottable/Afterglow.cpp
index 23c1858..ce72e9c 100644
--- a/render_2D/plottable/Afterglow.cpp
+++ b/render_2D/plottable/Afterglow.cpp
@@ -4,7 +4,6 @@
 #include "../render/Blend2D_Cache.h"
 #include "../renderable/Render_Partition.h"
 #include <algorithm>
-#include <array>
 #include <deque>
 namespace renderive {
 namespace detail {
@@ -66,31 +65,39 @@ void Afterglow_Control::publish() {
 }
 
 void Afterglow_Control::build_paint_task_graph(Renderable_Task_Graph& graph) {
-    const auto prepare = graph.emplace([this](const Scene_Render_Context&) {
-        prepare_render_frame(render_state_view());
+    auto& output = impl_->render_frame;
+    const int partition_count = output.partitioner.graph_partition_count(
+        render_properties(render_state_view()).partition_count.get(),
+        static_cast<int>(Scene_Base::task_executor_worker_count()));
+    const auto prepare = graph.emplace([this, partition_count](const Scene_Render_Context&) {
+        prepare_render_frame(render_state_view(), partition_count);
     }, "prepare afterglow");
-    std::array<Renderable_Task_Graph::Task, maximum_render_partitions> accumulation;
-    std::array<Renderable_Task_Graph::Task, maximum_render_partitions> coloring;
-    for (int index = 0; index < maximum_render_partitions; ++index) {
-        accumulation[static_cast<std::size_t>(index)] = graph.emplace(
+    std::vector<Renderable_Task_Graph::Task> accumulation;
+    accumulation.reserve(static_cast<std::size_t>(partition_count));
+    for (int index = 0; index < partition_count; ++index) {
+        const auto task = graph.emplace(
             [this, index](const Scene_Render_Context&) {
                 accumulate_partition(render_state_view(), index);
             },
-            "accumulate afterglow partition");
-        graph.precede(prepare, accumulation[static_cast<std::size_t>(index)]);
+            "accumulate afterglow partition " + std::to_string(index + 1));
+        graph.precede(prepare, task);
+        accumulation.push_back(task);
     }
     const auto normalize = graph.emplace([this](const Scene_Render_Context&) {
         normalize_render_frame();
     }, "normalize afterglow");
     for (const auto task : accumulation)
         graph.precede(task, normalize);
-    for (int index = 0; index < maximum_render_partitions; ++index) {
-        coloring[static_cast<std::size_t>(index)] = graph.emplace(
+    std::vector<Renderable_Task_Graph::Task> coloring;
+    coloring.reserve(static_cast<std::size_t>(partition_count));
+    for (int index = 0; index < partition_count; ++index) {
+        const auto task = graph.emplace(
             [this, index](const Scene_Render_Context&) {
                 color_partition(render_state_view(), index);
             },
-            "color afterglow partition");
-        graph.precede(normalize, coloring[static_cast<std::size_t>(index)]);
+            "color afterglow partition " + std::to_string(index + 1));
+        graph.precede(normalize, task);
+        coloring.push_back(task);
     }
     const auto compose = add_paint_task(
         graph, "compose afterglow",
@@ -99,7 +106,8 @@ void Afterglow_Control::build_paint_task_graph(Renderable_Task_Graph& graph) {
         graph.precede(task, compose);
 }
 
-void Afterglow_Control::prepare_render_frame(const Render_State_View& view) {
+void Afterglow_Control::prepare_render_frame(const Render_State_View& view,
+                                             int graph_partition_count) {
     const auto& state = render_properties(view);
     const auto& history = view.get(impl_->history);
     auto& output = impl_->render_frame;
@@ -125,8 +133,8 @@ void Afterglow_Control::prepare_render_frame(const Render_State_View& view) {
     output.work_size = static_cast<std::size_t>(width) * height;
     output.intensity.resize(output.work_size);
     output.pixels.resize(output.work_size);
-    output.active_partitions =
-        output.partitioner.begin(state.partition_count.get(), output.work_size);
+    output.active_partitions = output.partitioner.begin(graph_partition_count,
+                                                        output.work_size);
     output.valid = true;
 }
 
@@ -189,7 +197,12 @@ void Afterglow_Control::paint(Painter& painter, const Render_State_View& view) {
         return;
     painter.heatmap(output.layout.target, output.layout.width, output.layout.height,
                     output.pixels, Image_Interpolation_Mode::Bilinear);
-    output.partitioner.finish(output.work_size);
+    const auto& state = render_properties(view);
+    if (output.partitioner.finish(
+            state.partition_count.get(), output.active_partitions,
+            static_cast<int>(Scene_Base::task_executor_worker_count()),
+            output.work_size))
+        task_graph_changed();
 }
 }
 }
diff --git a/render_2D/plottable/Afterglow.h b/render_2D/plottable/Afterglow.h
index 4e31c54..daf5163 100644
--- a/render_2D/plottable/Afterglow.h
+++ b/render_2D/plottable/Afterglow.h
@@ -34,7 +34,7 @@ protected:
 private:
     struct Impl;
     std::unique_ptr<Impl> impl_;
-    void prepare_render_frame(const Render_State_View& state);
+    void prepare_render_frame(const Render_State_View& state, int graph_partition_count);
     void accumulate_partition(const Render_State_View& state, int partition_index);
     void normalize_render_frame();
     void color_partition(const Render_State_View& state, int partition_index);
diff --git a/render_2D/plottable/Plottable.h b/render_2D/plottable/Plottable.h
index f53d63a..6e5f4ab 100644
--- a/render_2D/plottable/Plottable.h
+++ b/render_2D/plottable/Plottable.h
@@ -60,6 +60,15 @@ public:
     template <auto Member, Property_Member_Assignable<Properties, Member> Value>
     Plottable_State& set(Value&& value) {
         Base::template set<Member>(std::forward<Value>(value));
+        if constexpr (requires { &Properties::partition_count; }) {
+            if constexpr (std::same_as<decltype(Member),
+                                       decltype(&Properties::partition_count)>) {
+                if constexpr (Member == &Properties::partition_count) {
+                    this->task_graph_changed();
+                    return *this;
+                }
+            }
+        }
         this->changed();
         return *this;
     }
diff --git a/render_2D/plottable/Spectrum.cpp b/render_2D/plottable/Spectrum.cpp
index b6b8a94..045a4b5 100644
--- a/render_2D/plottable/Spectrum.cpp
+++ b/render_2D/plottable/Spectrum.cpp
@@ -4,7 +4,6 @@
 #include "../render/Blend2D_Cache.h"
 #include "../renderable/Render_Partition.h"
 #include <algorithm>
-#include <array>
 #include <cmath>
 #include <iomanip>
 #include <sstream>
@@ -25,7 +24,7 @@ struct Spectrum_Interaction {
 using Spectrum_Interaction_State = Double_State_Strategy<Spectrum_Interaction_Base, Spectrum_Interaction>;
 struct Spectrum_Render_Frame {
     Adaptive_Render_Partitioner partitioner;
-    std::array<Blend2D_Color_Cache, maximum_render_partitions> layers;
+    std::vector<Blend2D_Color_Cache> layers;
     int active_partitions{1};
     std::size_t work_size{};
     bool valid{};
@@ -244,17 +243,24 @@ void Spectrum_Control::publish() {
 }
 
 void Spectrum_Control::build_paint_task_graph(Renderable_Task_Graph& graph) {
-    const auto prepare = graph.emplace([this](const Scene_Render_Context&) {
-        prepare_render_frame(render_state_view());
+    auto& output = impl_->render_frame;
+    const int partition_count = output.partitioner.graph_partition_count(
+        render_properties(render_state_view()).partition_count.get(),
+        static_cast<int>(Scene_Base::task_executor_worker_count()));
+    output.layers.resize(static_cast<std::size_t>(partition_count));
+    const auto prepare = graph.emplace([this, partition_count](const Scene_Render_Context&) {
+        prepare_render_frame(render_state_view(), partition_count);
     }, "prepare spectrum");
-    std::array<Renderable_Task_Graph::Task, maximum_render_partitions> partitions;
-    for (int index = 0; index < maximum_render_partitions; ++index) {
-        partitions[static_cast<std::size_t>(index)] = graph.emplace(
+    std::vector<Renderable_Task_Graph::Task> partitions;
+    partitions.reserve(static_cast<std::size_t>(partition_count));
+    for (int index = 0; index < partition_count; ++index) {
+        const auto partition = graph.emplace(
             [this, index](const Scene_Render_Context&) {
                 render_partition(render_state_view(), index);
             },
-            "paint spectrum partition");
-        graph.precede(prepare, partitions[static_cast<std::size_t>(index)]);
+            "paint spectrum partition " + std::to_string(index + 1));
+        graph.precede(prepare, partition);
+        partitions.push_back(partition);
     }
     const auto compose = add_paint_task(
         graph, "compose spectrum",
@@ -263,7 +269,8 @@ void Spectrum_Control::build_paint_task_graph(Renderable_Task_Graph& graph) {
         graph.precede(task, compose);
 }
 
-void Spectrum_Control::prepare_render_frame(const Render_State_View& view) {
+void Spectrum_Control::prepare_render_frame(const Render_State_View& view,
+                                            int graph_partition_count) {
     const auto& state = render_properties(view);
     const auto& published_frame = view.get(impl_->frame);
     auto& output = impl_->render_frame;
@@ -271,8 +278,8 @@ void Spectrum_Control::prepare_render_frame(const Render_State_View& view) {
     const Axis_Transform power_axis = impl_->power_axis->transform(view);
     output.valid = axes_are_orthogonal(frequency_axis, power_axis);
     output.work_size = published_frame ? published_frame->samples.size() : 0;
-    output.active_partitions =
-        output.partitioner.begin(state.partition_count.get(), output.work_size);
+    output.active_partitions = output.partitioner.begin(graph_partition_count,
+                                                        output.work_size);
     for (int index = 0; index < output.active_partitions; ++index)
         output.layers[static_cast<std::size_t>(index)].clear();
 }
@@ -380,7 +387,11 @@ void Spectrum_Control::paint(Painter& painter, const Render_State_View& view) {
             painter.text({box.x + 4.0, box.y + 3.0}, text.str(), state.tooltip_font, state.tooltip_text_pen);
         }
     }
-    output.partitioner.finish(output.work_size);
+    if (output.partitioner.finish(
+            state.partition_count.get(), output.active_partitions,
+            static_cast<int>(Scene_Base::task_executor_worker_count()),
+            output.work_size))
+        task_graph_changed();
 }
 }
 }
diff --git a/render_2D/plottable/Spectrum.h b/render_2D/plottable/Spectrum.h
index 2d9378e..a342067 100644
--- a/render_2D/plottable/Spectrum.h
+++ b/render_2D/plottable/Spectrum.h
@@ -65,7 +65,7 @@ protected:
 private:
     struct Impl;
     std::unique_ptr<Impl> impl_;
-    void prepare_render_frame(const Render_State_View& state);
+    void prepare_render_frame(const Render_State_View& state, int graph_partition_count);
     void render_partition(const Render_State_View& state, int partition_index);
     void publish() override;
 };
diff --git a/render_2D/plottable/Waterfall.cpp b/render_2D/plottable/Waterfall.cpp
index 57aa686..3a13198 100644
--- a/render_2D/plottable/Waterfall.cpp
+++ b/render_2D/plottable/Waterfall.cpp
@@ -3,7 +3,6 @@
 #include "Plottable_Real_Time_Data.h"
 #include "../renderable/Render_Partition.h"
 #include <algorithm>
-#include <array>
 #include <deque>
 #include <iomanip>
 #include <sstream>
@@ -98,17 +97,23 @@ void Waterfall_Control::publish() {
 }
 
 void Waterfall_Control::build_paint_task_graph(Renderable_Task_Graph& graph) {
-    const auto prepare = graph.emplace([this](const Scene_Render_Context&) {
-        prepare_render_frame(render_state_view());
+    auto& output = impl_->render_frame;
+    const int partition_count = output.partitioner.graph_partition_count(
+        render_properties(render_state_view()).partition_count.get(),
+        static_cast<int>(Scene_Base::task_executor_worker_count()));
+    const auto prepare = graph.emplace([this, partition_count](const Scene_Render_Context&) {
+        prepare_render_frame(render_state_view(), partition_count);
     }, "prepare waterfall");
-    std::array<Renderable_Task_Graph::Task, maximum_render_partitions> partitions;
-    for (int index = 0; index < maximum_render_partitions; ++index) {
-        partitions[static_cast<std::size_t>(index)] = graph.emplace(
+    std::vector<Renderable_Task_Graph::Task> partitions;
+    partitions.reserve(static_cast<std::size_t>(partition_count));
+    for (int index = 0; index < partition_count; ++index) {
+        const auto partition = graph.emplace(
             [this, index](const Scene_Render_Context&) {
                 render_partition(render_state_view(), index);
             },
-            "raster waterfall partition");
-        graph.precede(prepare, partitions[static_cast<std::size_t>(index)]);
+            "raster waterfall partition " + std::to_string(index + 1));
+        graph.precede(prepare, partition);
+        partitions.push_back(partition);
     }
     const auto compose = add_paint_task(
         graph, "compose waterfall",
@@ -117,7 +122,8 @@ void Waterfall_Control::build_paint_task_graph(Renderable_Task_Graph& graph) {
         graph.precede(partition, compose);
 }
 
-void Waterfall_Control::prepare_render_frame(const Render_State_View& view) {
+void Waterfall_Control::prepare_render_frame(const Render_State_View& view,
+                                             int graph_partition_count) {
     const auto& state = render_properties(view);
     const auto& rows = view.get(impl_->rows);
     auto& output = impl_->render_frame;
@@ -150,8 +156,8 @@ void Waterfall_Control::prepare_render_frame(const Render_State_View& view) {
     output.source_height = height;
     output.work_size = static_cast<std::size_t>(width) * height;
     output.pixels.resize(output.work_size);
-    output.active_partitions =
-        output.partitioner.begin(state.partition_count.get(), output.work_size);
+    output.active_partitions = output.partitioner.begin(graph_partition_count,
+                                                        output.work_size);
     output.valid = true;
 }
 
@@ -181,7 +187,11 @@ void Waterfall_Control::paint(Painter& painter, const Render_State_View& view) {
         return;
     painter.heatmap(output.layout.target, output.layout.width, output.layout.height,
                     output.pixels, state.interpolation_mode);
-    output.partitioner.finish(output.work_size);
+    if (output.partitioner.finish(
+            state.partition_count.get(), output.active_partitions,
+            static_cast<int>(Scene_Base::task_executor_worker_count()),
+            output.work_size))
+        task_graph_changed();
     if(state.tooltip_enabled && interaction.tooltip.active && output.layout.target.contains(interaction.tooltip.position)) {
         const Axis_Transform frequency_axis = impl_->frequency_axis->transform(view);
         const double frequency = frequency_axis.point_to_coord(interaction.tooltip.position);
diff --git a/render_2D/plottable/Waterfall.h b/render_2D/plottable/Waterfall.h
index 2123ea0..197a249 100644
--- a/render_2D/plottable/Waterfall.h
+++ b/render_2D/plottable/Waterfall.h
@@ -41,7 +41,7 @@ protected:
 private:
     struct Impl;
     std::unique_ptr<Impl> impl_;
-    void prepare_render_frame(const Render_State_View& state);
+    void prepare_render_frame(const Render_State_View& state, int graph_partition_count);
     void render_partition(const Render_State_View& state, int partition_index);
     void publish() override;
 };
diff --git a/render_2D/renderable/Render_Partition.h b/render_2D/renderable/Render_Partition.h
index 7113339..11f3c62 100644
--- a/render_2D/renderable/Render_Partition.h
+++ b/render_2D/renderable/Render_Partition.h
@@ -3,12 +3,10 @@
 #include <algorithm>
 #include <chrono>
 #include <cstddef>
-#include <thread>
+#include <limits>
 
 namespace renderive::detail {
 
-inline constexpr int maximum_render_partitions = 16;
-
 struct Render_Partition_Range {
     std::size_t first{};
     std::size_t last{};
@@ -25,45 +23,48 @@ inline Render_Partition_Range render_partition_range(std::size_t size,
 
 class Adaptive_Render_Partitioner {
 public:
-    Adaptive_Render_Partitioner() noexcept
-        : automatic_count_(std::clamp(
-              static_cast<int>(std::thread::hardware_concurrency()), 1,
-              maximum_render_partitions)) {}
-
-    int begin(int configured_count, std::size_t work_size) noexcept {
-        automatic_ = configured_count == 0;
-        const int requested = automatic_ ? automatic_count_ : configured_count;
-        const int useful = static_cast<int>(
-            std::min<std::size_t>(maximum_render_partitions,
-                                  std::max<std::size_t>(1, work_size)));
-        active_count_ = std::clamp(requested, 1,
-                                   std::min(maximum_render_partitions, useful));
-        started_at_ = Clock::now();
-        return active_count_;
+    [[nodiscard]] int graph_partition_count(int configured_count,
+                                            int worker_count) noexcept {
+        if (configured_count > 0)
+            return configured_count;
+        if (automatic_count_ == 0)
+            automatic_count_ = std::max(1, worker_count);
+        automatic_count_ = std::clamp(automatic_count_, 1,
+                                      std::max(1, worker_count));
+        return automatic_count_;
     }
 
-    void finish(std::size_t work_size) noexcept {
-        if (!automatic_)
-            return;
+    int begin(int graph_partition_count, std::size_t work_size) noexcept {
+        const int useful = static_cast<int>(std::min<std::size_t>(
+            static_cast<std::size_t>(std::numeric_limits<int>::max()),
+            std::max<std::size_t>(1, work_size)));
+        started_at_ = Clock::now();
+        return std::clamp(graph_partition_count, 1, useful);
+    }
+
+    [[nodiscard]] bool finish(int configured_count, int active_count,
+                              int worker_count, std::size_t work_size) noexcept {
+        if (configured_count != 0)
+            return false;
         const auto elapsed =
             std::chrono::duration_cast<std::chrono::microseconds>(Clock::now() - started_at_);
         constexpr auto target = std::chrono::microseconds(1500);
-        if (elapsed > target * 2 && active_count_ < maximum_render_partitions &&
-            work_size / static_cast<std::size_t>(active_count_) >= 512) {
-            automatic_count_ = std::min(maximum_render_partitions, active_count_ + 1);
-        } else if (elapsed < target / 2 && active_count_ > 1) {
-            automatic_count_ = active_count_ - 1;
-        } else {
-            automatic_count_ = active_count_;
-        }
+        const int previous = automatic_count_;
+        const int maximum = std::max(1, worker_count);
+        if (elapsed > target * 2 && active_count < maximum &&
+            work_size / static_cast<std::size_t>(active_count) >= 512)
+            automatic_count_ = std::min(maximum, active_count + 1);
+        else if (elapsed < target / 2 && active_count > 1)
+            automatic_count_ = active_count - 1;
+        else
+            automatic_count_ = active_count;
+        return automatic_count_ != previous;
     }
 
 private:
     using Clock = std::chrono::steady_clock;
     Clock::time_point started_at_{};
-    int automatic_count_{1};
-    int active_count_{1};
-    bool automatic_{true};
+    int automatic_count_{};
 };
 
 } // namespace renderive::detail
diff --git a/render_2D/renderable/Renderable.cpp b/render_2D/renderable/Renderable.cpp
index 602d803..c60c5a3 100644
--- a/render_2D/renderable/Renderable.cpp
+++ b/render_2D/renderable/Renderable.cpp
@@ -88,6 +88,11 @@ void Renderable::changed() noexcept {
     plot_.notify_model_dirty();
 }
 
+void Renderable::task_graph_changed() noexcept {
+    rebuild_task_graph();
+    plot_.notify_model_dirty();
+}
+
 Size Renderable::viewport_size() const noexcept {
     return plot_.viewport_size();
 }
diff --git a/render_2D/renderable/Renderable.h b/render_2D/renderable/Renderable.h
index c415aef..39bc8ca 100644
--- a/render_2D/renderable/Renderable.h
+++ b/render_2D/renderable/Renderable.h
@@ -50,6 +50,7 @@ protected:
                                                std::string name,
                                                Paint_Task_Function function);
     void changed() noexcept;
+    void task_graph_changed() noexcept;
     [[nodiscard]] Size viewport_size() const noexcept;
 
 private:
diff --git a/render_2D/tests/render_2D_Integration_Tests.cpp b/render_2D/tests/render_2D_Integration_Tests.cpp
index 08dc810..c94bdcb 100644
--- a/render_2D/tests/render_2D_Integration_Tests.cpp
+++ b/render_2D/tests/render_2D_Integration_Tests.cpp
@@ -652,7 +652,7 @@ TEST(Renderive_Core2, AxisRasterLayoutCoversAxisSwapAndEveryReversal) {
     EXPECT_DOUBLE_EQ(points.back().x, 130.0);
     EXPECT_DOUBLE_EQ(points.back().y, 30.0);
 }
-TEST(Renderive_Core2, PartitionedPlotsBuildExplicitInternalTaskGraphs) {
+TEST(Renderive_Core2, PartitionedPlotsBuildPropertyDrivenTaskGraphs) {
     Plot_Core plot;
     plot.init();
     const auto root = plot.root_renderable();
@@ -680,9 +680,26 @@ TEST(Renderive_Core2, PartitionedPlotsBuildExplicitInternalTaskGraphs) {
     EXPECT_EQ(spectrum->get<&Spectrum::Properties::partition_count>(), 4);
     EXPECT_EQ(waterfall->get<&Waterfall::Properties::partition_count>(), 4);
     EXPECT_EQ(afterglow->get<&Afterglow::Properties::partition_count>(), 4);
-    EXPECT_EQ(spectrum->task_graph()->nodes().size(), 18u);
-    EXPECT_EQ(waterfall->task_graph()->nodes().size(), 18u);
-    EXPECT_EQ(afterglow->task_graph()->nodes().size(), 35u);
+    EXPECT_EQ(spectrum->task_graph()->nodes().size(), 6u);
+    EXPECT_EQ(waterfall->task_graph()->nodes().size(), 6u);
+    EXPECT_EQ(afterglow->task_graph()->nodes().size(), 11u);
+
+    spectrum->set<&Spectrum::Properties::partition_count>(1);
+    waterfall->set<&Waterfall::Properties::partition_count>(1);
+    afterglow->set<&Afterglow::Properties::partition_count>(1);
+    spectrum->scene().publish_frame_state();
+    EXPECT_EQ(spectrum->task_graph()->nodes().size(), 3u);
+    EXPECT_EQ(waterfall->task_graph()->nodes().size(), 3u);
+    EXPECT_EQ(afterglow->task_graph()->nodes().size(), 5u);
+
+    spectrum->set<&Spectrum::Properties::partition_count>(0);
+    waterfall->set<&Waterfall::Properties::partition_count>(0);
+    afterglow->set<&Afterglow::Properties::partition_count>(0);
+    spectrum->scene().publish_frame_state();
+    const auto workers = Scene_Base::task_executor_worker_count();
+    EXPECT_EQ(spectrum->task_graph()->nodes().size(), workers + 2u);
+    EXPECT_EQ(waterfall->task_graph()->nodes().size(), workers + 2u);
+    EXPECT_EQ(afterglow->task_graph()->nodes().size(), workers * 2u + 3u);
 }
 TEST(Renderive_Core2, PlottableAxesAreDataDependenciesAndPaintOverlays) {
     Plot_Core plot;
diff --git a/web_server/app/Gallery_Observer_Adminive.h b/web_server/app/Gallery_Observer_Adminive.h
index d2cd0c8..42ffaf5 100644
--- a/web_server/app/Gallery_Observer_Adminive.h
+++ b/web_server/app/Gallery_Observer_Adminive.h
@@ -81,7 +81,21 @@ struct Type_Descriptor<renderive::Frame_Observer_Snapshot> {
             ADMINIVE_FIELD_LABEL(T, render_state_wait_ns, "Render State 等待"),
             ADMINIVE_FIELD_LABEL(T, render_finish_state_wait_ns, "Render Finish 等待"),
             ADMINIVE_FIELD_LABEL(T, queue_wait_ns, "回放队列等待"),
-            ADMINIVE_FIELD_LABEL(T, end_to_end_ns, "端到端延迟"))
+            ADMINIVE_FIELD_LABEL(T, end_to_end_ns, "端到端延迟"),
+            ADMINIVE_FIELD_LABEL(T, task_execution_state, "任务执行状态")
+                .enum_label<Task_Execution_State::Idle>("空闲")
+                .enum_label<Task_Execution_State::Queued>("已排队")
+                .enum_label<Task_Execution_State::Running>("执行中")
+                .enum_label<Task_Execution_State::Completed>("已完成")
+                .enum_label<Task_Execution_State::Failed>("失败"),
+            ADMINIVE_FIELD_LABEL(T, task_worker_count, "Taskflow 工作线程"),
+            ADMINIVE_FIELD_LABEL(T, task_graph_node_count, "任务图节点"),
+            ADMINIVE_FIELD_LABEL(T, task_completed_node_count, "已完成节点"),
+            ADMINIVE_FIELD_LABEL(T, task_running_node_count, "运行中节点"),
+            ADMINIVE_FIELD_LABEL(T, task_pending_node_count, "待执行节点"),
+            ADMINIVE_FIELD_LABEL(T, task_peak_parallelism, "峰值并行度"),
+            ADMINIVE_FIELD_LABEL(T, task_failed_node_count, "失败节点"),
+            ADMINIVE_FIELD_LABEL(T, task_graph_duration_ns, "任务图耗时"))
             .label("Kernel Observer");
     }
 };
diff --git a/web_server/app/Gallery_Protocol.cpp b/web_server/app/Gallery_Protocol.cpp
index 90fa97f..fb24065 100644
--- a/web_server/app/Gallery_Protocol.cpp
+++ b/web_server/app/Gallery_Protocol.cpp
@@ -85,7 +85,7 @@ adminive::Table_View action_view();
 namespace adminive {
 
 template <>
-struct Type_Descriptor<renderive::web::gallery_detail::Case_Model> {
+struct <renderive::web::gallery_detail::Case_Model> {
     static auto get() {
         using T = renderive::web::gallery_detail::Case_Model;
         return object<T>("renderive_gallery_case",
@@ -312,6 +312,20 @@ Json dashboard_contract() {
                 std::move(point_pair), std::move(bottleneck),
                 described_dashboard_field<renderive::Frame_Observer_Snapshot>(
                     "kernel_observer", "last_event"),
+                described_dashboard_field<renderive::Frame_Observer_Snapshot>(
+                    "kernel_observer", "task_execution_state"),
+                described_dashboard_field<renderive::Frame_Observer_Snapshot>(
+                    "kernel_observer", "task_worker_count"),
+                described_dashboard_field<renderive::Frame_Observer_Snapshot>(
+                    "kernel_observer", "task_graph_node_count"),
+                described_dashboard_field<renderive::Frame_Observer_Snapshot>(
+                    "kernel_observer", "task_running_node_count"),
+                described_dashboard_field<renderive::Frame_Observer_Snapshot>(
+                    "kernel_observer", "task_pending_node_count"),
+                described_dashboard_field<renderive::Frame_Observer_Snapshot>(
+                    "kernel_observer", "task_peak_parallelism"),
+                described_dashboard_field<renderive::Frame_Observer_Snapshot>(
+                    "kernel_observer", "task_graph_duration_ns"),
                 dashboard_field("像素超时", "client_performance.frame_request_timeout_count"),
                 dashboard_field("最近像素龄", "client_performance.last_pixel_receive_age_ms", "milliseconds", 0)
             })}
diff --git a/web_server/app/Gallery_Renderables.h b/web_server/app/Gallery_Renderables.h
index 289af82..97e111e 100644
--- a/web_server/app/Gallery_Renderables.h
+++ b/web_server/app/Gallery_Renderables.h
@@ -348,7 +348,8 @@ struct Type_Descriptor<renderive::web::Gallery_Time_Axis> {
 
 RENDERIVE_GALLERY_DESCRIPTOR(Gallery_Spectrum, Spectrum_Properties, "spectrum", "Spectrum",
     property_number<T, &P::frequency_point_size>("frequency_point_size", "Frequency points"),
-    property_number<T, &P::partition_count>("partition_count", "Render partitions"),
+    property_number<T, &P::partition_count>("partition_count", "Render partitions")
+        .description("0 = automatic, 1 = single task, N = N parallel partitions"),
     property_object<T, &P::frequency_range>("frequency_range", "Frequency range"),
     property_number<T, &P::center_frequency>("center_frequency", "Center frequency"),
     property_object<T, &P::sweep_frequency_range>("sweep_frequency_range", "Sweep range"),
@@ -378,7 +379,8 @@ RENDERIVE_GALLERY_DESCRIPTOR(Gallery_Waterfall, Waterfall_Properties, "waterfall
     property_object<T, &P::frequency_range>("frequency_range", "Frequency range"),
     property_object<T, &P::power_range>("power_range", "Power range"),
     property_number<T, &P::frequency_bin_count>("frequency_bin_count", "Frequency bins"),
-    property_number<T, &P::partition_count>("partition_count", "Render partitions"),
+    property_number<T, &P::partition_count>("partition_count", "Render partitions")
+        .description("0 = automatic, 1 = single task, N = N parallel partitions"),
     property_boolean<T, &P::visible_range_only>("visible_range_only", "Visible range only"),
     property_select<T, &P::interpolation_mode>("interpolation_mode", "Interpolation"),
     property_select<T, &P::color_map>("color_map", "Color map"),
@@ -392,7 +394,8 @@ RENDERIVE_GALLERY_DESCRIPTOR(Gallery_Afterglow, Afterglow_Properties, "afterglow
     property_object<T, &P::power_range>("power_range", "Power range"),
     property_number<T, &P::frequency_point_size>("frequency_point_size", "Frequency points"),
     property_number<T, &P::power_point_size>("power_point_size", "Power points"),
-    property_number<T, &P::partition_count>("partition_count", "Render partitions"),
+    property_number<T, &P::partition_count>("partition_count", "Render partitions")
+        .description("0 = automatic, 1 = single task, N = N parallel partitions"),
     property_boolean<T, &P::interpolate>("interpolate", "Interpolate power"),
     property_number<T, &P::attenuation_rate>("attenuation_rate", "Attenuation"),
     property_select<T, &P::color_map>("color_map", "Color map"));
diff --git a/web_server/tests/Web_Bridge_Tests.cpp b/web_server/tests/Web_Bridge_Tests.cpp
index a0bb4d8..503f1cf 100644
--- a/web_server/tests/Web_Bridge_Tests.cpp
+++ b/web_server/tests/Web_Bridge_Tests.cpp
@@ -232,7 +232,7 @@ TEST(RenderiveWebGallery, AdminiveCatalogCoversEveryControlCaseAndThreeModes) {
 TEST(RenderiveWebGallery, CatalogOwnsDashboardFieldsAndDynamicBehavior) {
     const nlohmann::json catalog = parse_json(Gallery_Protocol::catalog_json());
     const auto& dashboard = catalog.at("dashboard");
-    EXPECT_EQ(dashboard.at("performance").at("fields").size(), 15U);
+    EXPECT_EQ(dashboard.at("performance").at("fields").size(), 22U);
     EXPECT_EQ(dashboard.at("limits").at("fields").size(), 4U);
     EXPECT_EQ(dashboard.at("observer").at("sections").size(), 4U);
     std::set<std::string> client_sources;
@@ -291,6 +291,13 @@ TEST(RenderiveWebGallery, ObserverTelemetryAndDashboardUseAdminiveDescriptors) {
     for (const auto& field : feedback_descriptor.at("fields"))
         EXPECT_TRUE(feedback.contains(field.at("name").get<std::string>()));
     EXPECT_EQ(observer.at("mode"), "manual");
+    EXPECT_GT(observer.at("task_worker_count").get<std::size_t>(), 0U);
+    EXPECT_TRUE(observer.contains("task_execution_state"));
+    EXPECT_TRUE(observer.contains("task_graph_node_count"));
+    EXPECT_TRUE(observer.contains("task_running_node_count"));
+    EXPECT_TRUE(observer.contains("task_pending_node_count"));
+    EXPECT_TRUE(observer.contains("task_peak_parallelism"));
+    EXPECT_TRUE(observer.contains("task_graph_duration_ns"));
     EXPECT_FALSE(telemetry.contains("low_latency_limit"));
 }
 
@@ -344,6 +351,13 @@ TEST(RenderiveWebGallery, AdminiveResourcesPatchTheRealRenderableState) {
     EXPECT_EQ(spectrum.at("descriptor").at("name"), "spectrum");
     EXPECT_TRUE(spectrum.at("data").at("frequency_range").is_object());
     EXPECT_EQ(spectrum.at("data").at("frequency_point_size"), 512);
+    const auto partition_field = std::find_if(
+        spectrum.at("descriptor").at("fields").begin(),
+        spectrum.at("descriptor").at("fields").end(),
+        [](const auto& field) { return field.at("name") == "partition_count"; });
+    ASSERT_NE(partition_field, spectrum.at("descriptor").at("fields").end());
+    EXPECT_EQ(partition_field->at("presentation").at("description"),
+              "0 = automatic, 1 = single task, N = N parallel partitions");
 
     response = response_json(session.handle(gallery_request(
         Gallery_Request_Kind::Patch,
@@ -367,6 +381,39 @@ TEST(RenderiveWebGallery, AdminiveResourcesPatchTheRealRenderableState) {
     EXPECT_EQ(foreign.at("type"), "error");
 }
 
+TEST(RenderiveWebGallery, PartitionControlRebuildsTheObservedTaskGraph) {
+    Gallery_Plot_Session session;
+    ASSERT_EQ(response_json(session.handle(gallery_request(
+                  Gallery_Request_Kind::Open,
+                  open_message("spectrum", "manual")))).at("type"),
+              "case_state");
+
+    const auto render = [&session] {
+        EXPECT_EQ(invoke_action(session, "mode_prepare").at("type"), "case_state");
+        EXPECT_EQ(invoke_action(session, "mode_refresh").at("type"), "case_state");
+        return invoke_action(session, "mode_render").at("telemetry").at("kernel_observer");
+    };
+
+    ASSERT_EQ(patch_controls(session, {{"partition_count", 1}}, "spectrum").at("type"),
+              "case_state");
+    const auto single = render();
+    EXPECT_EQ(single.at("task_execution_state"), "Completed");
+    EXPECT_EQ(single.at("task_completed_node_count"),
+              single.at("task_graph_node_count"));
+    EXPECT_EQ(single.at("task_running_node_count"), 0);
+    EXPECT_EQ(single.at("task_pending_node_count"), 0);
+
+    ASSERT_EQ(patch_controls(session, {{"partition_count", 4}}, "spectrum").at("type"),
+              "case_state");
+    const auto partitioned = render();
+    EXPECT_EQ(partitioned.at("task_graph_node_count").get<std::size_t>(),
+              single.at("task_graph_node_count").get<std::size_t>() + 3U);
+    EXPECT_EQ(partitioned.at("task_completed_node_count"),
+              partitioned.at("task_graph_node_count"));
+    EXPECT_GT(partitioned.at("task_peak_parallelism").get<std::size_t>(), 0U);
+    EXPECT_GT(partitioned.at("task_graph_duration_ns").get<std::uint64_t>(), 0U);
+}
+
 TEST(RenderiveWebGallery, AxisSwapAndEveryPlotDirectionAreEditableAndRenderable) {
     const auto patch = [](Gallery_Plot_Session& session, std::string_view target,
                           nlohmann::json value) {