Make static actor Iris replication failure non-fatal during WP cell unload

Development Programming & Scripting Multiplayer & Networking
, ,
1

We found an replication issue described as follow, not sure if Epic has already found and fixed it:

When a client unloads a World Partition cell while the server has in-flight

replication data for a static actor in that cell, the client fails to resolve

the actor by path. Iris treats this as a permanent fatal error, disabling

replication for the object forever. This is a race condition with a narrow

window that triggers occasionally during normal gameplay or server mesh migration.

The error log is:

[2026.03.26-10.38.57:155][356]LogIris: Error: RepBridge(0)::Client cannot replicate object: RootObject XXXXX_UAID_025041000001B4C402_1507117266 (InternalIndex: 486) (0:148356077440549:0). Reported by client: ConnectionId:4 ViewTarget: ACTR_XXXXXX_C_2147480445 Named: [UNetConnection] RemoteAddr: XXXXXX, Name: XXXXX, Driver: Name:GameNetDriver Def:GameNetDriver XXXXXX_2147481265, IsServer: YES, PC: BP_XXXXX_PlayerController_C_2147479682, Owner: BP_XXXXX_PlayerController_C_2147479682, UniqueId: INVALID
[2026.03.26-10.38.57:155][356]LogOutputDevice: Error: === Handled ensure: ===
[2026.03.26-10.38.57:155][356]LogOutputDevice: Error: 
[2026.03.26-10.38.57:155][356]LogOutputDevice: Error: Ensure condition failed: false  [File:.\Runtime\Net\Iris\Private\Iris\ReplicationSystem\ObjectReplicationBridge.cpp] [Line: 3090] 
[2026.03.26-10.38.57:155][356]LogOutputDevice: Error: Client disabled replication for RootObject XXXXX_UAID_025041000001B4C402_1507117266  (InternalIndex: 486) (0:148356077440549:0). Client was unable to read the creation data and did not spawn the netobject.
[2026.03.26-10.38.57:155][356]LogOutputDevice: Error: Stack: 
[2026.03.26-10.38.57:155][356]LogOutputDevice: Error:

Fix: add a StaticActorLevelNotLoaded result flag that propagates through

NetActorFactory -> ObjectReplicationBridge -> ReplicationReader. The reader

skips the batch non-fatally (same path as FoundActorIsAuthorityAfterMigrating)

instead of sending ReplicationDisabled to the server. The server’s level group

filter will stop further sends once the client confirms the cell unload.

[Attachment Removed]

2

Hi,

This is a known issue, although I can’t provide any estimate as to when it may be fixed.

There was a change made in 5.7 (CL 43609783) related to this that removes the destroyed object’s handle from the list of handles to stop replicating, since the object may be respawned later. However, the ensure is still present.

We have an open internal task to address this, and I’ve included the info provided here on that tracker.

Thanks,

Alex

[Attachment Removed]

3

Hi Alex,

Thanks for the update.

So currently we’ll try to fix it by ourselves and later if Epic fixed it, we could integrate it into our repo. :+1:

[Attachment Removed]

4

Hi,

If you do make a fix, we’d be happy to look at a PR as well.

Thanks,

Alex

[Attachment Removed]