Flatpak fails to create GBM buffer on Nvidia Driver 580 (GTX 1050 Max-Q)

this story was created with the help of Google Gemini

**System Information**

OS: Linux Mint
GPU: NVIDIA GeForce GTX 1050 3GB Max-Q
Driver: nvidia-driver-580 (Version 580.173.02-0ubuntu0.24.04.1)
Package: Flatpak (io.github.CyberTimon.RapidRAW)
Memory: 8GB RAM / 6GB Swap (NVMe)

**Bug Description**
The application fails to initialize the graphical interface under the proprietary Nvidia driver 580. The log registers the activation of the "Nvidia explicit-sync workaround", followed by repeated GBM buffer creation failures and invalid arguments.

Steps to Reproduce

Install nvidia-driver-580 on Linux Mint.
Run the Flatpak version of RapidRAW via terminal.
Observe the crash output regarding GBM buffers.

**Standard Execution Logs**

> Gtk-Message: 09:32:15.268: Failed to load module "xapp-gtk3-module"
> Set ORT_DYLIB_PATH to: /app/lib/RapidRAW/resources/libonnxruntime.so
> [INFO] Applied Nvidia explicit-sync workaround (hardware compositing kept).
> [INFO] Found 2 cooperative matrix configurations supported by wgpu
> [INFO] Found 2 cooperative matrix configurations supported by wgpu
> Failed to create GBM buffer of size 1280x720: Argumento inválido
> Failed to create GBM buffer of size 1280x720: Argumento inválido

**Troubleshooting Attempted**

Isolating Wayland: Running flatpak run --nosocket=wayland io.github.CyberTimon.RapidRAW results in the exact same GBM buffer error.

Forcing OpenGL Backend: Running with --env=WGPU_BACKEND=gl crashes the initialization due to lack of compute shader support, dropping the GPU process entirely:

> [ERROR] indirect-validation error: ComputePipeline(Internal("The selected version doesn't support Features(BUFFER_STORAGE | COMPUTE_SHADER | DYNAMIC_ARRAY_SIZE)"))
> [WARN] GPU pre-initialization failed (editing and thumbnails may be degraded): Parent device is lost

Start by reproducing the Flatpak launch on Linux Mint with nvidia-driver-580 and capture the standard logs. Compare normal execution with --nosocket=wayland and --env=WGPU_BACKEND=gl, paying attention to the GBM buffer and compute-pipeline failures. Done means the graphical interface initializes under the reported configuration without the repeated GBM errors or GPU process loss.

Written by the indexing model from the issue text.