Rendering using the Workbench engine using Blender as a Python module on Amazon Linux 2023 causes libEGL error and takes way too long

I am trying to get Blender cloud rendering with Nvidia GPUs to work. I am using Blender 4.1 as a module. When rendering using Workbench (or Eevee) I get an error, the render takes much much longer than it should (50 sec vs 2 sec), and after the render is done I get a segmentation fault. Here is the error before the render: libEGL warning: egl: failed to create dri2 screen libEGL warning: egl: failed to create dri2 screen libEGL warning: failed to open /dev/dri/card1: Permission denied EGL Error (0x3009): EGL_BAD_MATCH: Arguments are inconsistent (for example, a valid context requires buffers not supplied by a valid surface). The Nvidia drivers should be setup correctly, running nvidia-smi outputs this: +-----------------------------------------------------------------------------------------+ | NVIDIA-SMI 570.124.06 Driver Version: 570.124.06 CUDA Version: 12.8 | |-----------------------------------------+------------------------+----------------------+ | GPU Name Persistence-M | Bus-Id Disp.A | Volatile Uncorr. ECC | | Fan Temp Perf Pwr:Usage/Cap | Memory-Usage | GPU-Util Compute M. | | | | MIG M. | |=========================================+========================+======================| | 0 NVIDIA A10G On | 00000000:00:1E.0 Off | 0 | | 0% 21C P8 15W / 300W | 1MiB / 23028MiB | 0% Default | | | | N/A | +-----------------------------------------+------------------------+----------------------+ +-----------------------------------------------------------------------------------------+ | Processes: | | GPU GI CI PID Type Process name GPU Memory | | ID ID Usage | |=========================================================================================| | No running processes found | +-----------------------------------------------------------------------------------------+ Any ideas what could be the problem here?