Large published island: server stops replicating a few seconds after a player enters - all versions, all regions, since v42.00

Summary

Since the v42.00 update, my published island is completely unplayable. Players reach the world, then the server stops replicating within seconds and everyone is disconnected with “Network Connection Lost” (errors.com.epicgames.common.processing) after the standard 90 s timeout. All published/private versions are affected, including ones that worked the day before on v41.30. An empty project on the same account, PC and region launches and runs perfectly, so this is specific to this (large) island.

Please select what you are reporting on:

Unreal Editor for Fortnite

What Type of Bug are you experiencing?

Stability

Steps to Reproduce

  1. Join island 7373-2475-5181 in Fortnite (any region — tested EU and Middle East).
  2. Matchmaking succeeds and the world loads; the Fortnite loading screen is dismissed.
  3. Stay in the session and do nothing.
  4. Within ~3 seconds the server stops replicating (client log: “Client timing out (no server replication in 2.00 seconds)”).
  5. After ~90 seconds the client is disconnected: “Network Connection Lost (errors.com.epicgames.common.processing)”.
  6. Repeat with private versions 16, 17, 18 and 19 — all fail identically.
  7. Launch an empty UEFN project on the same account/PC/region - it runs perfectly.

Expected Result

The island should run normally, as it did on v41.30 the day before with 138 concurrent players. The server should keep replicating and players should be able to play.

Observed Result

The server stops replicating a few seconds after a player enters and never recovers; the client is disconnected 90 s later.

Client log (FortniteGame.log):
15:47:12 Fortnite loading screen dismissed, player in world
15:47:17 LogFort: Warning: Client timing out (no server replication in 2.00 seconds)
15:48:45 LogNet: Warning: UNetConnection::Tick: Connection TIMED OUT. Elapsed: 90.02 Threshold: 90.00
LogNet: Error: FailureType = ConnectionTimeout
LogNet: Error: FailureType = ConnectionLost

The client stays healthy throughout (UI animations keep playing, camera still moves) — only the server goes silent.

On Middle East the server repeatedly stalls and RECOVERS instead of dying once:
16:41:06 / 16:41:09 / 16:41:16 no server replication (2.00 s each)
16:42:21 Network lag has been detected: GameNetDriver

A thread stuck in a loop never resumes. This one resumes and stalls again, producing frames longer than 2 seconds — the server cannot keep up rather than being deadlocked.

UEFN Launch Session reproduces it: the cook completes (“LoadingNewContent - Complete”) and the editor’s Valkyrie beacon then times out at its own 30 s threshold. In one session the server died 2 s after Start Game; in another it died ~19 s after the cook, with no island Verse having executed at all.

Platform(s)

PC (Windows). Reported by other players on their own platforms as well.

Island Code

7373-2475-5181

Additional Notes

Build: ++Fortnite+Release-42.00-CL-56878558. Worked normally on v41.30 (peak of 138 concurrent players two days ago). Project Size: 61 %.

Already ruled out, with log evidence:

  • Content beacon: “DownloadContent - 12s 316ms - Completed”.
  • Matchmaking/travel: PreTravel and PostTravel both signalled, “All members signaled”, “ClientJoin Completed (Duration=22.873)”, “MatchmakingGroup [shutdown]”. The join itself succeeds — the failure is after it, which is why this is filed under Stability rather than Matchmaking.
  • World Partition: “WorldPartition initialize took 806 us”.
  • Network version mismatch: client and UEFN both report NetCL 56183233, EngineNetworkVersion 45, Checksum 3858602336 - identical.
  • Account / PC / install: an empty project (Project Size 0 %) runs perfectly at 198 FPS on the same account, PC and region.
  • Region: reproduced on both EU and Middle East.
  • A specific published version: private versions 16-19 all fail identically.
  • Island Verse logic: I instrumented all 52 devices with a Print at the top of each OnBegin plus a 4 Hz heartbeat. In the session where the server died before Start Game, none of them printed - the island’s Verse never ran and the server died anyway.

Questions:

  1. Did v42.00 change server-side cost or any per-server limit for large islands?
  2. Is there any way for a creator to see why a session server dies? There is currently no server log, no error and no diagnostic - the only observable is that replication stops.

UPDATE — narrowed down to the LEVEL itself.

I created two fresh projects and migrated content into them:

  1. TestMapMigrate — migrated everything INCLUDING the level asset
    → reproduces the bug exactly.

  2. testMigratWithoutlvl — migrated all the same assets (Verse, VAT static
    meshes, textures, devices, materials) but kept the DEFAULT EMPTY LEVEL
    → works perfectly, no server hang.

So the failure is carried by the level asset, not by the Verse code, the assets
or the project size.

Memory is also ruled out: the Memory Calculation tool reports
Highest Memory Usage: 33,366 / 100,000 — only 33 % of budget.

The level uses World Partition on a ~14.8 km island with runtime data layers
driven by level sequences (Material Element Switcher tracks, ~513 tracks per
sequence). This all worked on v41.30 the day before.

FORT-1143204 has been created and its status is ‘Unconfirmed’. This is now in a queue to be reproduced and confirmed.

I’m having this exact issue and my island isn’t even particularly large, it is predominantly scene graph-based though. When I delete a chunk of the entities it does load, but I get what appear to be multi-second lag spikes and ultimately kicked.

On the occasions where I’m kicked out during the connecting… phase, my log tells me the island isn’t downloaded.

[2026.08.20-21.40.51:401][597]MatchmakingLog: [2a02] Updating local graceful upgrade state based on Wahooney's selection: 
		Island: 
		LinkId [Mnemonic=[playlist_vk_edit] Version=[latest]] 
		Session [Session Id [None] Join Info [IslandJoinability [CanNotBeJoinedOrWatched] SessionKey [None]]] 
		MatchmakingSettings (V1) [
			WorldId [World [None] WorldName [None] WorldOwnerId [INVALID] IsJoinable [Unjoinable]] 
			ProductModes [None] 
			Privacy [Fill] 
			Region [EU]]
		Timestamp: 2026.08.20-21.40.50 Unix timestamp: 1787262050
		Using graceful upgrade: True
		Id: (redacted)
[2026.08.20-21.40.51:401][597]LogMatchmakingUtility: [3273] LinkId: 'Mnemonic=[playlist_vk_edit] Version=[latest]'.
[2026.08.20-21.40.51:401][597]SelectedIslandViewModelLog: Skipping island's settings button and download progress refresh as not in the front end
[2026.08.20-21.40.51:401][597]MatchmakingLog: [2837] Not matchmaking : island is not downloaded