Compatibility Report: ARC Raiders (1808500) — Reproducible GameThread Hang Returning to Speranza

System Information

Component | Configuration -- | -- Distribution | Linux Mint 22.x (Ubuntu 24.04 base) Kernel | 6.8.0-147-generic Desktop | Cinnamon / Xorg GPU | NVIDIA GeForce RTX 3080 Ti, 12 GB VRAM NVIDIA drivers tested | 580.178.04 and 595.99.02 System RAM | 64 GB installed CPU | 32 logical processors Steam AppID | 1808500 Proton versions tested | Hotfix, Experimental, Experimental bleeding-edge

These crashes occurred despite changes to NVIDIA drivers, Proton versions, and synchronization settings. No GPU crash was identified by Unreal in these reports.

Live GDB Backtrace Investigation

Two separate GDB thread backtraces were collected while the game was frozen, before the process terminated.

Files:

• arc-hang-backtrace.txt — captured under Proton Hotfix

• arc-hang-backtrace-2.txt — captured under Proton Experimental

The captures contain approximately 207 and 215 Linux threads, respectively.

Both captures include threads executing Wine synchronization functions such as:

syscall()
fsync_wait_objects()
NtWaitForSingleObject()
__wine_syscall_dispatcher()

Other threads show:

NtWaitForMultipleObjects()
NtWaitForAlertByThreadId()
NtDelayExecution()
NtRemoveIoCompletion()
NtRemoveIoCompletionEx()

The captures also include ordinary Wine thread-pool waits and Vulkan-related activity.

GDB reported the following warnings:

Selected architecture i386:x86-64 is not
compatible with reported target architecture i386:x64-32
Unable to find libthread_db matching inferior's

thread library, thread debugging will not be available.

Because of these debugger limitations and the lack of symbolized PioneerGame functions, these thread stacks do not conclusively identify the object being waited on, the responsible thread, or whether the root cause is inside Proton or the game itself.

The key observation is that Unreal independently and repeatedly detects the same GameThread hang during a specific transition.

Troubleshooting Performed

The following changes were individually tested without eliminating the recurring hang:

NVIDIA

• NVIDIA 580.178.04

• NVIDIA 595.99.02

• GPU recognized correctly by nvidia-smi

• No NVIDIA Xid errors found in the inspected kernel logs surrounding earlier failures

Proton

• Proton Hotfix

• Proton Experimental

• Proton Experimental bleeding-edge

• Tested disabling FSYNC and ESYNC with:

PROTON_NO_FSYNC=1 PROTON_NO_ESYNC=1 %command%

The same Unreal hang remained reproducible.

Steam and Game Environment

• Disabled Steam Overlay

• Shader pre-caching already enabled

• Verified game-file integrity through Steam

• Backed up the existing Proton compatibility prefix and allowed Proton to recreate a fresh prefix

• Launched and played with the regenerated prefix

• Reproduced the same return-to-Speranza hang afterward

Logging and Crash Analysis

• Reviewed Denuvo Anti-Cheat runtime logs from earlier failures

• Examined Unreal's CrashContext.runtime-xml files

• Examined Linux coredumpctl information from earlier segmentation faults

• Collected two live GDB thread backtraces

Earlier sessions also experienced SIGSEGV crashes. However, the consistently reproducible transition failure documented here is Unreal's GameThread Hang, which does not necessarily have the same root cause as those earlier segmentation faults.

Additional Observations

• Gameplay during raids generally runs normally.

• The failure is associated specifically with leaving a completed raid.

• The game freezes before the post-raid summary screen.

• Unreal reports GPUCRASH_NONE.

• The same PioneerGame stack offsets recur across independent crashes.

• The issue persists with two NVIDIA driver versions.

• The issue persists across several Proton configurations.

• Rebuilding the Proton prefix did not resolve the failure.

Attachments

Two GDB backtraces are available:

1. arc-hang-backtrace.txt

2. arc-hang-backtrace-2.txt

I can also provide the relevant Unreal CrashContext.runtime-xml files.

Request to Proton Maintainers

Could this reproducible return-to-Speranza GameThread hang be investigated for a possible Proton/Wine compatibility issue?

In particular, is there a known synchronization, rendering-transition, or game-runtime behavior that could explain the main game thread failing to progress during the transition from a completed raid back to Speranza?

The available evidence does not conclusively establish a Wine deadlock or GPU driver fault.

I'm happy to provide additional targeted diagnostics if the existing Unreal crash contexts and GDB backtraces are insufficient.

Related issue: #9164 — general ARC Raiders compatibility tracking. This report concerns a specific repeatable transition hang rather than the original out-of-video-memory failure.