Getting Xid 109 hangs on my end even with proton-cachyos-slr(20260703).
So far I've narrowed it down to whether HDR is enabled or not.

With xwayland (HDR not enabled) I can get to the brightness select screen on initial startup.
With Wayland + HDR enabled it always freezes when trying to load the license screen.
With gamescope + --hdr-enabled same deal as with Wayland + HDR.

RTX 5070ti
Driver 615.71.09 (open module)

XIDs for me with ray tracing enabled and ultra settings on driver 615 with a 5090. It works if I use VKD3D_CONFIG=descriptor_heap or downgrade to 595 without that though. Ultra settings, gamescope with HDR, highest as raytracing will allow.

Follow-up (RTX 5090, driver 615.71.09, CachyOS, KDE Wayland).

Correction to my report: I said the game works on proton-cachyos-slr 20260703. It loads, but it isn't immune: I got two Xid 109 hangs in gameplay around the first combat ability (about 35 min and about 4 min into two sessions), then a clean 43-minute session. Those report Info 0x2c013, versus 0x17c013 / 0x15c011 on the load-time hangs.

VKD3D_CONFIG=descriptor_heap looks like a fix for the load-time hang on GE-Proton11-7:

• Without it, GE-Proton11-7 gave Xid 109 on the loading screen on 4 of 5 launches (the fifth ran about 2 min with no Xid; I don't remember whether it loaded). Proton Experimental hung on its 1 launch.
• With it: 3 of 3 launches loaded past that point with no Xid, and the third has run 16+ minutes so far. The log shows VK_EXT_descriptor_heap enabled. There is a noticeable stutter on the loading screen each time.
• Small sample, so suggestive rather than proven. It matches matte-schwartz's report.

HDR: with HDR settings unchanged, adding the flag stopped the hang, so HDR by itself isn't sufficient to cause it on my setup. All of these runs were native Wayland, with HDR enabled in KWin and requested in Proton (PROTON_ENABLE_WAYLAND=1 PROTON_ENABLE_HDR=1 ENABLE_HDR_WSI=1 DXVK_HDR=1), including the 43-minute cachyos session and the three descriptor_heap loads. kareigu's result on a 5070 Ti may hold on that setup.

Hung-run log: 17.6 s after vkd3d-proton init, 244x vkd3d_allocate_device_memory: Memory allocation failed, falling back to system memory in about 1.2 s, then VK_ERROR_DEVICE_LOST about 4.6 s later. None of that in the clean runs. VRAM in use was about 12.8 of 32 GB. The Proton Experimental hang log has no such lines, so this may not be the common factor.

Not yet tested: whether the flag also stops the gameplay hang on cachyos.

Hey! Just if it can be of help, I have another data point: VKD3D_CONFIG=descriptor_heap fixes it on an RTX 4070 Ti SUPER as well.

System

• GPU: NVIDIA GeForce RTX 4070 Ti SUPER (16 GB)
• Driver: 615.71.09 (open kernel module)
• OS: Bazzite 44 (KDE, Fedora 44 base), kernel 7.2.4
• Proton: Experimental bleeding-edge (experimental-bleeding-edge-11.0-440709-20260926)

Without the workaround: with ray tracing enabled (even on Low), the game hangs as soon as I click New Game, every time. The kernel logs Xid 109:

NVRM: Xid (PCI:0000:01:00): 109, pid=..., name=CONTROLResonant, channel 0x00000012, errorString CTX SWITCH TIMEOUT, Info 0x2cc019
NVRM: Xid (PCI:0000:01:00): 109, pid=..., name=CONTROLResonant, channel 0x00000012, errorString CTX SWITCH TIMEOUT, Info 0x6c019

vkd3d-proton then reports VK_ERROR_DEVICE_LOST. Things that did not help:

• older Proton Experimental and the bleeding-edge build from 2026-09-24
• PROTON_DISABLE_NVAPI=1 (no DLSS)
• VKD3D_CONFIG=breadcrumbs (no breadcrumb report was produced)

With VKD3D_CONFIG=descriptor_heap: the log shows VK_EXT_descriptor_heap enabled. The game starts a New Game with ray tracing on and plays normally, with no Xid and no DEVICE_LOST.

For the moment, I've had no trouble with Indirect Lightning.

I hope this can be of help! (:

Reproducible trigger + it's the Wayland driver, not HDR or RT (RTX 5070 Ti)

System: RTX 5070 Ti, driver 615.71.09 (open module), Bazzite 44.20260921 (kernel 7.2.4), KDE Plasma 6.7.5 Wayland, proton-cachyos 11.0-20260703-slr.

Repro (100% on my setup): Evacuation Zone, next to the Survivalist's Bunker, at the vending machine: dialogue "Cool robot" → "Draw". The game freezes on selecting "Draw" (audio keeps playing), then Xid 109 CTX SWITCH TIMEOUT.

All runs below had VKD3D_CONFIG=descriptor_heap, same Proton, same save:

So on my machine the native Wayland driver is the trigger, with or without HDR, with or without ray tracing. descriptor_heap fixed the hang on the loading screen for me but not this one.

VKD3D_CONFIG=breadcrumbs produced no breadcrumb report. The PROTON_LOG of one hung run shows VK_ERROR_DEVICE_LOST right after the Xid (vkd3d_wait_for_gpu_timeline_semaphore: Failed to wait ... vr 2). There are also 3× Memory allocation failed, falling back to system memory about 60 s earlier, probably unrelated.

I couldn't test Proton Experimental bleeding-edge (20260926) with Wayland: it doesn't honor PROTON_ENABLE_WAYLAND, and the game loaded winex11.

@JordHisu I was having a very similar issue: the game would crash whenever I did a few specific things like that. It started after the game's latest update. It went away after I deleted the shader cache folder.

@JordHisu Same behavior here on different hardware.

System: RTX 4090, driver 615.71.09, Ryzen 9 9950X3D, Gentoo with a custom 7.2.4 kernel, Wayland with HDR.

Trigger: picking a dialogue option in the West Incursion Zone. Same freeze on selecting the option, then Xid 109 CTX SWITCH TIMEOUT (Info 0xac01b on mine).

Same results as your table: with the native Wayland driver on, it hangs whether HDR and RT are on or off, and with PROTON_ENABLE_WAYLAND=0 it passes.

Playing past it now with RT on using:
VKD3D_CONFIG=descriptor_heap DXVK_HDR=1 PROTON_ENABLE_HDR=0 PROTON_ENABLE_WAYLAND=0 WINE_CPU_TOPOLOGY=8:0,1,2,3,4,5,6,7 gamemoderun %command%
(the CPU topology part is just an affinity tweak for the X3D chip, shouldn't be related)

System: RTX 4070 super, driver 615.71.09, 5800X3D, 7.2.4 kernel, Wayland.

Most projector videos are corrupted when using RT Direct Lighting.

Setting on High/Ultra:

Setting on Off:

This setting also breaks shadows in certain quest areas.

A lot of these are late in the game, haven't found anyone with similar issues yet.

System

GPU: NVIDIA GeForce RTX 4070 Ti SUPER (16 GB)
Driver: 615.71.09 (open kernel module)
OS: Bazzite 44 (KDE, Fedora 44 base), kernel 7.2.4
Proton: Experimental bleeding-edge (experimental-bleeding-edge-11.0-440709-20260926)

@1ukas I have the same problem with the projections, realised it in the exact same quest hehe (:

Follow-up (RTX 5090, driver 615.71.09 open module, CachyOS kernel 7.2.8, KDE Plasma 6.7.5 Wayland, HDR). Confirming @JordHisu's finding: the native Wayland driver is the trigger on my setup too.

I found a dialogue choice that hung every time with PROTON_ENABLE_WAYLAND=1 and passes with it off. All runs used VKD3D_CONFIG=descriptor_heap and the same save:

The Wayland runs went on → off → on → off with nothing else changed, and the hang followed the setting each time.

• Not a build or cache issue. The hang reproduced on stock GE-Proton11-7. The other three Wayland hangs were on a GE-Proton11-7 build with the RTX Mega Geometry branches (Saancreed/vkd3d-proton:nvapi-extras, jp7677/dxvk-nvapi:d3d12-extras). One of those came right after I deleted the Steam shader cache and vkd3d-proton.cache, so the cache fix @thebozzcl mentioned didn't help here.
• Not every dialogue. A different choice that hung once passed on the next 2 tries.
• Gamescope HDR is worse, not a workaround. It hung 3 of 3 launches, twice in menus within about 40 s of launch.
• Not only at dialogue. With Wayland on, I also had random hangs in menus and combat. Stock GE-Proton11-7 hung 3 times in about 20 min of play today.
• Possibly relevant (a guess, not verified): Wine's Wayland driver and the gamescope WSI layer both present through a Wayland surface. The only passing setup presents through X11/XWayland. That, together with @JordHisu's result that Wayland hangs with HDR off too, suggests the Wayland presentation path is the trigger rather than HDR.
• The PROTON_LOGs show nothing unusual before the hang, just signaller_worker timeouts and then VK_ERROR_DEVICE_LOST. I can attach any of them.

Same Xid 109 on an RTX 5080, but on proton-cachyos-slr (cachyos-11.0-20260703-slr), the build that works in the original report. In my case HDR appears to be the trigger.

System:

• GPU: RTX 5080
• Driver: 615.71.09 (open kernel module)
• CPU: Ryzen 7 9800X3D
• Kernel: CachyOS, KDE Plasma Wayland. First HDR hang on 7.2.7-1-cachyos, second on 7.2.9-1-cachyos

Results on proton-cachyos-slr, same prefix and driver, ray tracing off:

Symptoms: the game freezes with audio still playing. The game process is in R state, not D, and responds to kill -9 from a TTY. Time to hang varies from a few minutes to over an hour.

Xid lines from two HDR hangs (first with PROTON_USE_WAYLAND=1 DXVK_HDR=1, second with PROTON_ENABLE_WAYLAND=1 DXVK_HDR=1):

NVRM: Xid (PCI:0000:01:00): 109, pid=5601, name=CONTROLResonant, channel 0x00000013, errorString CTX SWITCH TIMEOUT, Info 0x21c01c
NVRM: Xid (PCI:0000:01:00): 109, pid=11175, name=CONTROLResonant, channel 0x0000001b, errorString CTX SWITCH TIMEOUT, Info 0x32c024

PROTON_LOG=1 log from the second hang is attached. No vkd3d, DEVICE_LOST or timeout errors are logged. The log goes quiet at the Xid and ends ~110 s later when the process was killed.

control-hdr-hang.log