[Debian]Bumblebee causes failure of nvidia-persistenced.service at startup

[samveen]

I have Debian Bookworm on my workstation (T14 Gen 2 with a TU117M [GeForce MX450]). nvidia-driver (debian::bookworm/non-free), firmware-nvidia-gsp (debian::bookworm/non-free-firmware) and bumblebee (debian::bookworm/main) are installed from Debian Bookworm repositories.

optirun/primusrun both run correctly (I'm running JA2 onwine-staging:i386 from winehq::bookworm/main)

DKMS status is good:

bbswitch/0.8, 6.1.0-34-amd64, x86_64: installed
nvidia-current/535.216.01, 6.1.0-34-amd64, x86_64: installed

Modules are loaded as expected:

$ sudo lsmod |grep -iE 'nvid|bb'
nvidia_uvm           1540096  0
nvidia_modeset       1314816  0
nvidia              56795136  17 nvidia_uvm,nvidia_modeset
bbswitch               20480  0
drm                   614400  12 drm_kms_helper,drm_display_helper,nvidia,drm_buddy,thinkpad_acpi,i915,ttm
video                  65536  3 thinkpad_acpi,i915,nvidia_modeset

As bumblebee modprobe config blacklists nvidia/nvidia-current , and then loads them after bbswitch is loaded, the creation of the /dev/nvidia* device nodes is delayed. This causes the nvidia-persistenced.service service to fail in a non-deterministic manner, whenever this service doesn't find the device nodes it expects with the error(in the journal):

nvidia-persistenced[1311]: Failed to query NVIDIA devices. Please ensure that the NVIDIA device files (/dev/nvidia*) exist, and that user 113 has read and write permissions for those files.

To make this issue go away without modifying the nvidia-* packages, one solution I can think off is to add a systemd device for /dev/nvidia0 and add an systemd add-in for nvidia-persistenced.service to wait for this device (that's what I'm implementing on my laptop.

Is there any better solution to fix this?