The deferred-release path was never wired up (BeginFrame's loop over PendingGPUReadbackBuffers is unreachable because EndFrame calls PendingGPUReadbackBuffers.Reset() before the next BeginFrame can see the entries). 

EnqueueReadback() itself does not free the offset either. It only uses the offset to read from the count buffer. As a result, every GPU readback path through an NDC Writer DI permanently leaks one entry from FNiagaraGPUInstanceCountManager. UsedInstanceCounts has no decrement path, so the stat "Used GPU Instance Counters" grows monotonically.

Removing the Water plugin eliminates the errors entirely. No water assets exist in the level — this is purely the precacher walking the plugin's registered VF permutations against materials in the cook set.

We've encountered a crash in UWidgetAnimation::CreateDirectorInstance when repeatedly switching resolutions. We've had it for a while but it's not until recently that we've found a consistent repro.

Our understanding is that it's a lifetime issue. Sequence of events:

1. Viewport resizing triggers Slate to rebuild its widgets
2. Some widget with an active animation gets destroyed/marked for delete
3. UMGSequenceTickManager::ForceFlush is called
4. During Flush MovieSceneEventSystem::TriggerEvents fires
5. Consequently CreateDirectorInstance is called on UWidgetAnimation
6. GetPlaybackContex() returns null
7. CastChecked LogFatal (crash)

PCG can be used to spawn Instanced Actors via the PCG IA Interop plugin, that makes Spawned Instanced Actors available as node. When a PCGActor placed in the level runs a PCG graph that spawns Instanced Actors, this creates Instanced Actors Managers with instances.

Problem

The created Instanced Actor Manager actors get MainGrid assigned as Runtime Grid, even if the PCGActor that spawned them didn’t have a Runtime Grid value (i.e. its None). Next time the map is loaded, this results in error messages:

Actor /Game/MAP_ReproPCG.PCG_PlaceInstancedActors references an actor in a different runtime grid /Game/MAP_ReproPCG.InstancedActorsManager_-1_-1_0
Actor /Game/MAP_ReproPCG.PCG_PlaceInstancedActors references an actor in a different runtime grid /Game/MAP_ReproPCG.InstancedActorsManager_0_-1_0
Actor /Game/MAP_ReproPCG.PCG_PlaceInstancedActors references an actor in a different runtime grid /Game/MAP_ReproPCG.InstancedActorsManager_-1_0_0
Actor /Game/MAP_ReproPCG.PCG_PlaceInstancedActors references an actor in a different runtime grid /Game/MAP_ReproPCG.InstancedActorsManager_0_0_0

Suggested Fix

The PCG actor and spawned IAMs runtime grids should match. If the PCG actor had it set to None, perhaps leave the IAMs value as None as well.

Instanced Actors can be moved by unchecking Lock Instance Locations on an IAM, and double-clicking into the actor → comp → instance to select an instance.

Problem

Moving an IA via the gizmo isn’t properly recorded as an undoable/redoable transaction. It can’t be undone or redone.

Suggested Fix

For user friendliness, the action should be undoable.

Actors can be converted into Instanced Actors via the IA plugin. During runtime and in the editor, an instanced actor’s visual is generated from the collection of StaticMeshComponents on the actor blueprint.

Problem

Any non-uniform transforms of StaticMeshComps on the actor blueprint aren’t correctly previewed in the editor. Instead, the IA representation in editor shows all meshes with the identity transform, albeit the IA’s location is correct. The relative transform from actor to mesh isn’t correctly previewed. See screenshots and repro project.

Suggested Fix

The relative transform should be correctly previewed.

Root cause is in UMovieSceneSkeletalAnimationSection::GetAutoSizeRange (MovieSceneSkeletalAnimationSection.cpp:337). It calls InnerToOuterTransform.TryTransformTime(InnerEndTime).Get(InnerEndTime), and when the inverse transform fails to map InnerEndTime to an outer time (because the value is outside the Inner Clamp or Inner Loop range), the .Get(InnerEndTime) fallback returns the raw inner-tick value. That value is then used as if it were an outer-tick value in GetInclusiveStartFrame() + (OuterEndTime - OuterStartTime).FrameNumber, mixing inner-tick and outer-tick values and producing a negative or otherwise invalid length. SetRange then refuses the range and Auto Size has no effect.

This behavior was introduced by CL 40829088 ( [Link Removed] Closed , fixed in 5.6), which replaced the previous TOptional check with a .Get(fallback) pattern.

Assert location: Plugins/Runtime/HairStrands/Source/HairStrandsCore/Private/GroomManager.cpp:2011

check(InstanceData.CardInstance->Guides.RestResource);

Root cause:
Card LOD guide resources are not created when only Groom Cache drives deformation (no simulation, graph deformer, global interpolation, or skinning binding).

In GroomComponent.cpp (~line 3219), bNeedDeformedPositions for Card LOD omits bIsSimulationCacheEnable:

const bool bNeedDeformedPositions = bHasSimulation || bHasDeformation || bHasGlobalDeformation || BindingType == EHairBindingType::Skinning;

When this is false, InstanceLOD.Guides.RestResource and related Card LOD guide resources are never assigned (~lines 3256–3277).

Proposed fix:
Add Groom Cache to Card LOD initialization in GroomComponent.cpp:

const bool bNeedDeformedPositions =
    bHasSimulation || bHasDeformation || bHasGlobalDeformation
    || BindingType == EHairBindingType::Skinning
    || HairGroupInstance->HairGroupPublicData->bIsSimulationCacheEnable;

Ensure condition failed: InstanceDesc.InstanceId <= 0xFFFFFF
[File: RayTracingInstanceBufferUtil.cpp] [Line: 316]
InstanceId must fit in 24 bits.

Root Cause

GetDynamicRayTracingInstances() in NiagaraRendererMeshes.cpp (line ~1660) force-disables sort/cull for the RT path:

ParticleMeshRenderData.bNeedsSort = false;
ParticleMeshRenderData.bNeedsCull = false;
ParticleMeshRenderData.bSortCullOnGpu = false;

This causes PerformSortAndCull() to return the total particle count for every mesh variant,. Each mesh variant creates an FRayTracingInstance with NumTransforms set to all particles. The total TLAS instances is NumParticles × NumMeshVariants. The raster path handles this correctly via SortAndCullIndices / FNiagaraSortCullHelper::IsVisibile() which filters by MeshIndex.

The InstanceId is set to InstanceSceneDataOffset + TransformIndex with bIncrementUserDataPerInstance = true (in PrimitiveSceneInfo.cpp:SetupCachedRayTracingInstance), which can exhausts the 24-bit DXR InstanceId space.

Additional Notes

• Path Tracer disables RT instance culling (RayTracingInstanceCulling.cpp:71), so all ghost instances are included in the TLAS.
• Temporal samples do NOT multiply the instance buffer — the TLAS is built once per frame.
• CalculateMeshUsed() already scans particles to determine which meshes are active, but only produces a boolean per variant, not a count. It also only works for CPU sims.

Potential Fix

Re-enable MeshIndex culling for the RT path. For CPU sims, the existing FNiagaraSortCullHelper already filters by MeshIndex — remove the force-disable in the RT path and call InitializeSortInfo before PerformSortAndCull. For GPU sims, a lightweight count pass or one-frame-latent readback of per-variant counts would be needed since particle data is not CPU-accessible.

Affected Code

• Engine/Plugins/FX/Niagara/Source/Niagara/Private/NiagaraRendererMeshes.cpp — GetDynamicRayTracingInstances()

Audio that uses an Attenuation to set Spatialized Audio (just positioned, nothing to do with HRTFs or multichannel plugins) does not get rendered with the MRQ files. Initial discovery revealed that the listener is not getting updated correctly in MRQ code.

Slack thread: [Link Removed]

Relates to EPS case : [Link Removed]

Relates to Git PR : [Link Removed]

Root cause is in UMovieSceneSkeletalAnimationSection::GetAutoSizeRange (MovieSceneSkeletalAnimationSection.cpp:337). It calls InnerToOuterTransform.TryTransformTime(InnerEndTime).Get(InnerEndTime), and when the inverse transform fails to map InnerEndTime to an outer time (because the value is outside the Inner Clamp or Inner Loop range), the .Get(InnerEndTime) fallback returns the raw inner-tick value. That value is then used as if it were an outer-tick value in GetInclusiveStartFrame() + (OuterEndTime - OuterStartTime).FrameNumber, mixing inner-tick and outer-tick values and producing a negative or otherwise invalid length. SetRange then refuses the range and Auto Size has no effect.

This behavior was introduced by CL 40829088 ( [Link Removed] Closed , fixed in 5.6), which replaced the previous TOptional check with a .Get(fallback) pattern.

See attached PDF for more details.