diff --git a/MobileGL/MG_Backend/DirectVulkan/Renderer/UniformManager.cpp b/MobileGL/MG_Backend/DirectVulkan/Renderer/UniformManager.cpp
index 41b9398b..b96aedab 100644
--- a/MobileGL/MG_Backend/DirectVulkan/Renderer/UniformManager.cpp
+++ b/MobileGL/MG_Backend/DirectVulkan/Renderer/UniformManager.cpp
@@ -650,6 +650,14 @@ namespace MobileGL::MG_Backend::DirectVulkan {
             return false;
         }
 
+        // The shader may write this buffer, and those writes land in GPU memory behind the
+        // frontend's CPU shadow - which is what MapBuffer and GetBufferSubData read.
+        // Host-visible coherent GPU residency makes the shadow BE that memory, so the
+        // results are visible without a readback path, exactly as for a capture buffer.
+        bufferObject->EnsureGpuResidentStorage();
+        // ... and the read that follows has to wait for this draw or dispatch to retire.
+        bufferObject->MarkGpuWritten();
+
         BufferSlice slice{};
         if (!m_bufferManager->AcquireResidentSlice(BufferKind::ShaderStorage, bufferObject, slice) || !slice.IsValid()) {
             MGLOG_E("ResolveStorageBufferDescriptor: failed to sync GL buffer %u for block '%s'",
diff --git a/MobileGL/MG_Backend/DirectVulkan/Renderer/VkBufferManager.cpp b/MobileGL/MG_Backend/DirectVulkan/Renderer/VkBufferManager.cpp
index 22297278..28a884cd 100644
--- a/MobileGL/MG_Backend/DirectVulkan/Renderer/VkBufferManager.cpp
+++ b/MobileGL/MG_Backend/DirectVulkan/Renderer/VkBufferManager.cpp
@@ -7,6 +7,8 @@
 // End of Source File Header
 
 #include "VkBufferManager.h"
+#include "../DirectVulkan.h"
+#include "VulkanRenderer.h"
 
 namespace MobileGL::MG_Backend::DirectVulkan {
     namespace {
@@ -57,6 +59,18 @@ namespace MobileGL::MG_Backend::DirectVulkan {
             }
         }
 
+        // The CPU is about to read a buffer a shader wrote. Its bytes live in coherent
+        // host-visible GPU storage (EnsureGpuResidentStorage adopts it when the buffer is
+        // bound as a shader storage buffer), so nothing needs copying - but coherence only
+        // says the writes are visible once they have happened, so the work has to retire
+        // first.
+        void Ops_ReadbackFromGpu(BufferObject& bufferObject) {
+            (void)bufferObject;
+            if (pVulkanRenderer) {
+                pVulkanRenderer->FinishPendingGpuWork();
+            }
+        }
+
         void* Ops_AcquirePersistentMap(BufferObject& bufferObject) {
             if (g_activeBufferManager) {
                 return g_activeBufferManager->AcquirePersistentMap(bufferObject);
@@ -80,6 +94,7 @@ namespace MobileGL::MG_Backend::DirectVulkan {
             .FlushMappedRange = Ops_FlushMappedRange,
             .OnDestroy = Ops_OnDestroy,
             .AcquirePersistentMap = Ops_AcquirePersistentMap,
+            .ReadbackFromGpu = Ops_ReadbackFromGpu,
         };
     } // namespace
 
diff --git a/MobileGL/MG_Backend/DirectVulkan/Renderer/VulkanRenderer.cpp b/MobileGL/MG_Backend/DirectVulkan/Renderer/VulkanRenderer.cpp
index 295389f9..d7303faf 100644
--- a/MobileGL/MG_Backend/DirectVulkan/Renderer/VulkanRenderer.cpp
+++ b/MobileGL/MG_Backend/DirectVulkan/Renderer/VulkanRenderer.cpp
@@ -6919,6 +6919,17 @@ void main() {
 
     }
 
+    Bool VulkanRenderer::FinishPendingGpuWork() {
+        auto& frame = m_frameContext.GetCurrent();
+        if (!frame.isCommandRecording) {
+            return true;
+        }
+        if (VkRenderPassManager::GetActiveRenderPass() != nullptr) {
+            VkRenderPassManager::EndRenderPass(frame.commandBuffer);
+        }
+        return SubmitReadbackCommandsAndWait(frame);
+    }
+
     Bool VulkanRenderer::SubmitReadbackCommandsAndWait(FrameContext::FrameData& frame) {
         if (frame.isCommandRecording) {
             m_frameContext.EndCommandRecording();
diff --git a/MobileGL/MG_Backend/DirectVulkan/Renderer/VulkanRenderer.h b/MobileGL/MG_Backend/DirectVulkan/Renderer/VulkanRenderer.h
index 4b0221f7..a6a80df3 100644
--- a/MobileGL/MG_Backend/DirectVulkan/Renderer/VulkanRenderer.h
+++ b/MobileGL/MG_Backend/DirectVulkan/Renderer/VulkanRenderer.h
@@ -795,6 +795,14 @@ namespace MobileGL::MG_Backend::DirectVulkan {
                                            VkImageLayout finalLayout);
         Bool SubmitReadbackCommandsAndWait(FrameContext::FrameData& frame);
 
+    public:
+        // Submits whatever is recorded and waits for it. The CPU is about to read memory
+        // a shader wrote (a mapped shader storage buffer), and coherent host-visible
+        // storage only guarantees visibility once the work that produced it has retired.
+        Bool FinishPendingGpuWork();
+
+    private:
+
         void ShutdownSwapchain();
 
         // Static functions
diff --git a/MobileGL/MG_State/GLState/BufferState/BufferObject.cpp b/MobileGL/MG_State/GLState/BufferState/BufferObject.cpp
index af971f34..c4ebb279 100644
--- a/MobileGL/MG_State/GLState/BufferState/BufferObject.cpp
+++ b/MobileGL/MG_State/GLState/BufferState/BufferObject.cpp
@@ -175,9 +175,6 @@ namespace MobileGL::MG_State::GLState {
     }
 
     void BufferObject::MarkGpuWritten() {
-        // A GPU-resident buffer has no separate shadow to refresh: reads already resolve
-        // against the coherent map the shader wrote into.
-        if (m_resource.IsGpuResident()) return;
         m_gpuWritePending = true;
     }
 
diff --git a/MobileGL/MG_State/GLState/BufferState/BufferObject.h b/MobileGL/MG_State/GLState/BufferState/BufferObject.h
index d6268970..c689614e 100644
--- a/MobileGL/MG_State/GLState/BufferState/BufferObject.h
+++ b/MobileGL/MG_State/GLState/BufferState/BufferObject.h
@@ -157,8 +157,10 @@ namespace MobileGL {
             void WritebackFromBackend(DataPtr data, SizeT atOffset);
 
             // A draw or dispatch just ran with this buffer bound where a shader can write
-            // it (shader storage / atomic counter): the GPU copy may now differ from the
-            // shadow, so the next read has to pull it back.
+            // it (shader storage / atomic counter). The next read has to reconcile with
+            // that: pull the bytes back, or - when the shadow already IS coherent GPU
+            // memory - wait for the work that wrote them to retire. Which of the two is
+            // the backend's business; the flag only says a GPU write is outstanding.
             void MarkGpuWritten();
             // Refreshes the shadow from the backend when a GPU write is outstanding. Called
             // from every path that reads the shadow on the app's behalf.