You need to unload and modprobe the nvidia modules again. Cannot exactly remember which are loaded by try:
lsmod|grep nvidia
and rmmod the modules and modprobe them again. Then try restarting bumblebee.
Maybe there is lacking some config to get them unloaded before suspend and loaded afterwards, but what can I say,
nvidia on Linux is not the most stable experience.
With suse-prime you may get a more stable experience, but I cannot promise anything…
You are not alone, that is the “standard” behaviour with bumblebee in my limited experience; but I must admit to never trying to debug that beyond removing and reloading the nvidia modules as bosim suggested.
On the other hand, with suse-prime I always have to switch to intel before shutting down, rebooting or suspending to ensure that the graphical target (AKA runlevel 5) resumes or reboots with no hiccups, but that might be HW dependent.
Makes sense then. I sort of assumed that you used the proprietary driver. When I had an optimus machine the chipset
was not supported (GM108), but I think it was in for kernel 4.7, so I was forced to use the proprietary software, and let me
say it in this way, it is not the most stable software in the world.