I would think that a 3-second delay would be absurdly bad for a video-based show,
especially if you're running audio over an external program, which would desync the webcams from the audio. You'll hear guests laughing at something, then see they react 3 seconds later, or hear them stop talking but their mouth keeps moving, or making gestures about something they were talking about 3 seconds ago. If you kept the webcams in sync with the audio, then you'll have the perpetual "satellite" problem of people responding 3 seconds after you ask a question. There will be that awkward silence while audio catches up between the two of you, and good luck trying to interrupt one another.
And beyond that, what if you want these multiple guests to be able to see and hear one another, too? You'd all have to be streaming to everyone else, all at the same time, all suffering from the delay problem. It would be a mess, and you'd have to use an external program anyway.
OBS is not video conferencing software, and it's not supposed to be. If you want to do video conferencing and include it in OBS, then use a video conferencing program. It's perfectly fine to use the correct programs for the correct jobs.
If you really want to try it, you would have to set up your own RTMP server and have people stream to that, and then bring in their streams with the Video source plugin. Here's a guide for setting up an RTMP server:
https://obsproject.com/forum/resources/how-to-set-up-your-own-private-rtmp-server-using-nginx.50/