Screen 2 will be running any apps I need to show (coding etc.) and screen 3 is at the back of the workbench, which has a downward facing camera so I can show me working on the boards and other electronics.
Switching scenes from main cam to what's on screen 2 to the workbench is very handy, even without a dedicated key pad.
The only hassle is that under Linux, OBS sometimes forgets the camera device designations on startup and I have to pull out the USB connections, re-plug and then go into both camera definitions and re-select the right cameras.
I have device video and audio sink definitions behind he scenes so that the video and audio can be fed to virtual devices inside Zoom or Teams etc.
Edit: Oh, and I have a client plugin which allows me to use my Android phone camera as a roving webcam to highlight things on the workbench such as a scope screen, power supply current reading etc.
All works quite well.
The Arch wiki has fairly in-depth instructions for understanding how udev rules work https://wiki.archlinux.org/title/udev
OBS scenes is the killer feature for me, so it just works well. In addition the AFK one I've also set a couple of zoom levels that frame me better if I decide to slide my seat back a bit.
I even have scenes for sprint planning estimation sessions. 1-3-5-8 etc superimposed over my picture. :D (they are numbered buttons on the Streamdeck).
I do love reading about all these setups though. I feel like so many tech companies and software engineers are new to remote work that these tips can go a long way to interfacing effectively with coworkers. At least in my case it was trial and error until I got what I wanted.
I also have a very similar setup. High quality webcam, high quality headset, and a few lights pointing into my face to offset the lighting in the room. I have Zoom configured to use original audio and turn off echo cancellation and most audio postprocessing as the headset pretty much does everything I need audio wise. I've even had my partner talking in the same room and others on my VC haven't heard a thing.
My webcam needs almost no touch-up. While I can stream at 1080p easily (and do so with friends over FFMpeg and an RTSP server), Zoom (and I'm guessing Meet) will frequently downsample your webcam output when in a VC with multiple participants to save on bandwidth. A 1:1 chat can have you streaming in full 1080p glory, but I don't ever need a post-processing layer atop my webcam.
The setup is as easy as me turning on my camera and putting on my headset. Make sure you are using Ethernet so you don't run into AP congestion, especially if you have others in the house doing VC as well. But yeah modern high quality VC is almost completely plug-and-play.