I tend to use the simplest, but most high-effort approach of "show all the people in the speaking". I'm personally not big on the tendency a lot of webcomics have to only show one character in most panels even when two or more people are conversing, partially because it's less clear, but partially because I think the expression or reaction of the listener can add valuable context to what the speaker is saying
If I can, I try to avoid deviating from "speaker on the left speaks first", but sometimes it's necessary to avoid breaking the 180 Degree Rule. If somebody speaks from off-panel, It's usually because they're far away and are about to enter the area, because if they are near enough to the other person, I'd put them on the panel, even if it's just the back of their head.
Chibi faces can work well in a lighter-toned comic, like say a shoujo manga where everyone's babbling away saying silly things, it would work great there, but it'd feel pretty jarring in more serious situations. Some people find coloured text hard to read (a widely known issue in the Homestuck fandom) and of course, since colour blindness is a thing, you should never over-rely on colour to differentiate things; there should always be some secondary cue; that's just good basic UX/UI design.
If in the thumbnails it looked like a page was likely to be a clusterbomb of confusing bubbles, I'd probably look at alternative ways to do that conversation to make it clearer, unless I wanted it to be a a confusing babble or voices on purpose.