diff --git a/core/libs/prx/libSceAgcDriver/Graphics/src/ShaderResources.cpp b/core/libs/prx/libSceAgcDriver/Graphics/src/ShaderResources.cpp
index f09d2cbcb66..d4c8100530d 100644
--- a/core/libs/prx/libSceAgcDriver/Graphics/src/ShaderResources.cpp
+++ b/core/libs/prx/libSceAgcDriver/Graphics/src/ShaderResources.cpp
@@ -616,10 +616,19 @@ std::shared_ptr<Texture> cachedTexture(const Context& context, std::span<const s
     // after them is current at the generation of a second (memoized) collect. The store marks the
     // surface's DCC keys uncompressed, so the keys are read after it.
     const auto flushStart = profile ? std::chrono::steady_clock::now() : std::chrono::steady_clock::time_point{};
-    if (source == nullptr && !clearThroughKeys && StorageTexture::FlushPending(address, bytes, nullptr, "sampled texture")) {
+    const auto observedKeys = keys.has_value() ? *keys : scanKeys();
+    const bool stagedClear = source != nullptr && resource.dccAddress != 0 && (IsDccClear(observedKeys) || observedKeys == DccKeys::ClearSingle);
+    bool stagedFlushed = false;
+    if ((source == nullptr || stagedClear) && !clearThroughKeys && StorageTexture::FlushPending(address, bytes, nullptr, "sampled texture")) {
         if (profile) LookupOutcomes::Add(LookupOutcomes::PendingFlush, flushStart);
         generation = GuestMemory::CollectWrites(address, bytes);
-        keys.reset();
+        if (stagedClear) {
+            stagedFlushed = true;
+            keys = DccKeys::Uncompressed;
+        } else {
+            keys.reset();
+        }
+        if (stagedFlushed) source.reset();
     }
     if (!keys.has_value()) keys = scanKeys();
     if (disabled) {
@@ -656,7 +665,7 @@ std::shared_ptr<Texture> cachedTexture(const Context& context, std::span<const s
             // The view follows the storage image, whatever the GPU wrote to it since; guest memory
             // written meanwhile is taken in by refreshing the image. A fast clear the image cannot
             // see (keys, and no pending results to prefer) ends the view: a snapshot holds the clear.
-            if ((source == nullptr || source == it->source) && (source != nullptr || *keys == DccKeys::Uncompressed) && StorageImageServesKeys(*it->source, resource.dccAddress)) {
+            if (!stagedFlushed && (source == nullptr || source == it->source) && (source != nullptr || *keys == DccKeys::Uncompressed) && StorageImageServesKeys(*it->source, resource.dccAddress)) {
                 if (!GuestMemory::UnchangedSince(address, bytes, it->source->Generation())) it->source->Refresh();
                 if (it->texture->IsDepthCopy()) it->texture->RefreshDepthCopy();
                 it->keys = *keys;
diff --git a/core/libs/prx/libSceAgcDriver/tests/execution/StagedClearSampledRead.cpp b/core/libs/prx/libSceAgcDriver/tests/execution/StagedClearSampledRead.cpp
new file mode 100644
index 00000000000..8997e9418ad
--- /dev/null
+++ b/core/libs/prx/libSceAgcDriver/tests/execution/StagedClearSampledRead.cpp
@@ -0,0 +1,208 @@
+#include "prx/libSceAgcDriver/Execution/include/GuestMemory.hpp"
+#include "prx/libSceAgcDriver/Execution/include/VulkanDevice.hpp"
+#include "prx/libSceAgcDriver/Graphics/include/DccMetadata.hpp"
+#include "prx/libSceAgcDriver/Graphics/include/ColorTargetLayout.hpp"
+#include "prx/libSceAgcDriver/Graphics/include/GuestTextureResource.hpp"
+#include "prx/libSceAgcDriver/Graphics/include/ShaderResources.hpp"
+#include "prx/libSceAgcDriver/Graphics/include/Texture.hpp"
+#include "prx/libSceAgcDriver/Graphics/include/TextureTiling.hpp"
+#include "prx/libc/include/GuestAllocations.hpp"
+#include "prx/libc/include/GuestArena.hpp"
+#include "prx/libc/include/GuestWriteWatch.hpp"
+#include "Recompiler.hpp"
+#include "VulkanTestDevice.hpp"
+#ifdef _WIN32
+#ifndef NOMINMAX
+#define NOMINMAX
+#endif
+#include <windows.h>
+#else
+#include <sys/mman.h>
+#endif
+#include <algorithm>
+#include <array>
+#include <cstdint>
+#include <cstdio>
+#include <cstdlib>
+#include <cstring>
+#include <iostream>
+#include <memory>
+#include <mutex>
+#include <span>
+#include <vector>
+
+namespace {
+
+using AgcDriver::Graphics::DccKeys;
+using AgcDriver::Graphics::Require;
+using AgcDriver::Graphics::StorageTexture;
+using ShaderRecompiler::ShaderStage;
+
+constexpr std::uint32_t Threads = 32;
+constexpr std::uint32_t Side = 128;
+constexpr std::uint32_t Format8888Unorm = 56;
+constexpr std::uint32_t TileRenderTarget64KB = 0x1b;
+constexpr std::uint32_t Type2D = 9;
+constexpr std::size_t SurfaceBytes = 65536;
+constexpr std::size_t KeyBytes = SurfaceBytes / 256;
+constexpr std::size_t KeysOffset = SurfaceBytes;
+constexpr std::size_t SampleTexelBytes = 4;
+constexpr std::size_t UnalignedBlockBytes = 2 * 65536 + 2048;
+constexpr std::size_t ArenaBytes = 3 * 65536;
+
+alignas(256) constexpr std::array<std::uint32_t, 8> WriteRowCode{
+    0x7e020281, 0x7e0402f2, 0x7e0602f2, 0x7e0802f2, 0x7e0a02f2, 0xf0201f08, 0x00010200, 0xbf810000,
+};
+
+alignas(256) constexpr std::array<std::uint32_t, 11> SampleSecondTexelCode{
+    0x34060084, 0x7e2802ff, 0x3c400000, 0x7e2a02ff, 0x3b800000, 0xf09c0f08, 0x00610a14, 0xbf8c3f70,
+    0xe0781000, 0x80000a03, 0xbf810000,
+};
+
+constexpr std::array<std::uint32_t, 4> PointSampler{0x00000092u, 0x00fff000u, 0u, 0u};
+
+std::uint64_t AddressOf(const void* data) {
+    return static_cast<std::uint64_t>(reinterpret_cast<std::uintptr_t>(data));
+}
+
+class GuestBlock {
+public:
+    GuestBlock() {
+#ifdef _WIN32
+        block = static_cast<std::uint8_t*>(GuestArena::GuestArenaAllocate_nid_postfix(ArenaBytes, 65536));
+        if (block != nullptr) GuestArena::GuestArenaCommit_nid_postfix(block, ArenaBytes, PAGE_READWRITE, ArenaBytes);
+#else
+        constexpr std::uintptr_t alignment = 65536;
+        void* mapped = mmap(nullptr, ArenaBytes + alignment, PROT_READ | PROT_WRITE, MAP_PRIVATE | MAP_ANONYMOUS, -1, 0);
+        if (mapped != MAP_FAILED) {
+            const auto begin = reinterpret_cast<std::uintptr_t>(mapped);
+            const auto aligned = (begin + alignment - 1) & ~(alignment - 1);
+            if (aligned != begin) munmap(mapped, aligned - begin);
+            if (aligned + ArenaBytes != begin + ArenaBytes + alignment) munmap(reinterpret_cast<void*>(aligned + ArenaBytes), begin + alignment - aligned);
+            block = reinterpret_cast<std::uint8_t*>(aligned);
+            GuestWriteWatch::GuestWriteWatchRegister_nid_postfix(block, ArenaBytes);
+        }
+#endif
+        Require(block != nullptr, "staged clear sampled read: cannot allocate the guest block");
+        Require(AgcDriver::GuestMemory::Watched(AddressOf(block), UnalignedBlockBytes), "staged clear sampled read: the guest block is not write-watched");
+        GuestAllocations::Mutation().Add(block, UnalignedBlockBytes, true, true, true);
+    }
+
+    ~GuestBlock() {
+        GuestAllocations::Mutation().Remove(block);
+#ifdef _WIN32
+        GuestArena::GuestArenaReset_nid_postfix(block, ArenaBytes);
+        GuestArena::GuestArenaRelease_nid_postfix(block, ArenaBytes);
+#else
+        munmap(block, ArenaBytes);
+        GuestWriteWatch::GuestWriteWatchUnregister_nid_postfix(block, ArenaBytes);
+#endif
+    }
+
+    GuestBlock(const GuestBlock&) = delete;
+    GuestBlock& operator=(const GuestBlock&) = delete;
+
+    std::uint8_t* Data() { return block; }
+
+private:
+    std::uint8_t* block = nullptr;
+};
+
+std::array<std::uint32_t, 8> TextureDescriptor(const std::uint8_t* texels, const std::uint8_t* keys) {
+    const auto address = AddressOf(texels);
+    const auto meta = AddressOf(keys) >> 8u;
+    return {
+        static_cast<std::uint32_t>(address >> 8u),
+        static_cast<std::uint32_t>((address >> 40u) & 0xffu) | (Format8888Unorm << 20u) | (((Side - 1u) & 3u) << 30u),
+        ((Side - 1u) >> 2u) | ((Side - 1u) << 14u),
+        0xfacu | (TileRenderTarget64KB << 20u) | (Type2D << 28u),
+        0u,
+        0u,
+        (1u << 20u) | (1u << 21u) | (static_cast<std::uint32_t>(meta & 0xffu) << 24u),
+        static_cast<std::uint32_t>(meta >> 8u),
+    };
+}
+
+std::array<std::uint32_t, 4> BufferDescriptor(const void* data, std::uint32_t bytes) {
+    const auto address = AddressOf(data);
+    return {static_cast<std::uint32_t>(address), static_cast<std::uint32_t>((address >> 32u) & 0xffffu), bytes, 0x31016facu};
+}
+
+void Dispatch(AgcDriver::VulkanDevice& device, std::span<const std::uint32_t> code, const std::vector<std::uint32_t>& userData) {
+    const std::array<ShaderRecompiler::MemoryRegion, 1> memory{{{reinterpret_cast<std::uintptr_t>(code.data()), std::as_bytes(code)}}};
+    const ShaderRecompiler::ShaderComputeStageInfo compute{{Threads, 1, 1}, 0u, {false, false, false}, false, 1};
+    ShaderRecompiler::RecompileRequest request{
+        {ShaderStage::Compute, reinterpret_cast<std::uintptr_t>(code.data()), code, 0, {}},
+        {32, 0, userData, compute, std::nullopt, std::nullopt, memory},
+        device.Target(),
+        {0, 0, 0, 128}
+    };
+    request.useCache = false;
+    const auto result = ShaderRecompiler::Recompile(request);
+    device.Dispatch(result, 1, 1, 1, {}, reinterpret_cast<std::uintptr_t>(code.data()));
+    device.WaitIdle();
+}
+
+std::size_t Count(const std::uint8_t* bytes, std::size_t size, std::uint8_t value) {
+    return static_cast<std::size_t>(std::count(bytes, bytes + size, value));
+}
+
+}
+
+int main() {
+    try {
+        const auto device = OpenVulkanTestDevice();
+        if (!device) return VulkanTestSkipped;
+        GuestBlock block;
+        auto* texels = block.Data();
+        auto* keys = texels + KeysOffset;
+        std::memset(texels, 0x00, SurfaceBytes);
+        const AgcDriver::Graphics::ColorTargetLayout layout(Side, Side, AgcDriver::Graphics::ColorTileMode::RenderTarget, 4);
+        std::memset(texels + layout.Offset(1, 0), 0xff, SampleTexelBytes);
+        std::memset(keys, 0xff, KeyBytes);
+        const auto surface = AddressOf(texels);
+        const auto texture = TextureDescriptor(texels, keys);
+        const auto resource = AgcDriver::Graphics::DecodeTextureResource(texture);
+        AgcDriver::GuestMemory::BumpCollectEpoch();
+        std::vector<std::uint32_t> userData(24, 0u);
+        std::vector<std::uint32_t> outputData(Threads * 4u, 0xdeadbeefu);
+        const auto results = BufferDescriptor(outputData.data(), Threads * 16u);
+        std::copy(results.begin(), results.end(), userData.begin());
+        std::copy(texture.begin(), texture.end(), userData.begin() + 4);
+        std::copy(PointSampler.begin(), PointSampler.end(), userData.begin() + 12);
+        {
+            std::lock_guard gpu(AgcDriver::GuestMemory::GpuMutex());
+            std::vector<std::uint32_t> writeData(16, 0u);
+            std::copy(texture.begin(), texture.end(), writeData.begin() + 4);
+            AgcDriver::GuestMemory::BumpCollectEpoch();
+            Dispatch(*device, WriteRowCode, writeData);
+            const auto image = StorageTexture::FindPending(surface, SurfaceBytes);
+            Require(image != nullptr, "the pass left no pending results at the surface");
+            Require(image->UploadedKeys() == DccKeys::Uncompressed, "the pass did not draw under the settled keys");
+            std::memset(keys, 0x00, KeyBytes);
+            AgcDriver::GuestMemory::BumpCollectEpoch();
+            Require(AgcDriver::Graphics::TextureClearKeys(resource, SurfaceBytes) == DccKeys::Clear0000, "the key bytes do not read a staged 0000 clear");
+            Dispatch(*device, SampleSecondTexelCode, userData);
+            Require(StorageTexture::FindPending(surface, SurfaceBytes) == nullptr, "the sampled read did not store the pending results: the lookup serves without them in guest memory");
+            const auto written = Count(texels, SurfaceBytes, 0xff);
+            const auto cleared = Count(texels, SurfaceBytes, 0x00);
+            char message[320];
+            std::snprintf(message, sizeof(message), "the sampled read stored %zu written and %zu cleared bytes of %zu, expected %u written", written, cleared, SurfaceBytes, Threads * 4u + SampleTexelBytes);
+            Require(written == Threads * 4u + SampleTexelBytes && cleared == SurfaceBytes - written, message);
+            for (std::uint32_t index = 0; index < Threads * 4u; ++index) {
+                std::snprintf(message, sizeof(message), "component %u of thread %u is 0x%08x, expected the drawn 0x3f800000 the staged clear does not cover", index % 4u, index / 4u, outputData[index]);
+                Require(outputData[index] == 0x3f800000u, message);
+            }
+        }
+        {
+            std::lock_guard lock(AgcDriver::GuestMemory::GpuMutex());
+            AgcDriver::Graphics::FlushCachedTextures(device->Device());
+        }
+        device->WaitIdle();
+        std::puts("staged clear sampled read tests passed");
+        return 0;
+    } catch (const std::exception& error) {
+        std::cerr << error.what() << '\n';
+        return 1;
+    }
+}