The issue I see is, that for cutscenes in BG3 to play out normally for players unengaged in the cutscene and for continuaty you request, they would need to be fully animated from top down view, would require for "actors" to stay and naturally move from positions they initated the conversation, and then the camera would need to procedurally capture the "footage". That I think in unachievable - really no hand scripting would be possible, unless like in Halo example, players would need to move into a very specific positions to engage in a conversation, and all NPC would be completely static. Otherwise everything would need to be generated - actors movements would need to be generated to respond to the location of the actors, and camera would need to be generated to capture what is happening.
Sorry, you're exagarating the complication here, while that puzzle was already solved long ago in DOS2 for co-op. In that game when you take a backseat in co-op for certain important dialogues the game takes a little pause and takes away control from you to make your character come to the needed point. For example you are perfectly safe to stay somewhere in corridor when conversation before the final boss-fight happens, your character will run inside the room and stop by side of another player character himself. Same goes for NPCs, sometimes they just take a moment at the start to move to needed coordinates. And BG3 certainly has "walk from wherever you're to this coordinates"-logic in it's engine, it's used for our movement all the time.
But with all that said I would be fine If Larian would just use a bit more "mirroring" scripts to at least mask the problem with semi-prerecorded cutscenes better. Because it's one thing when I switch to some companion and see that my character is standing still instead of playing some complicated animation, but it's totally another and much worse thing when I see how Wyll runs into the room after the cutscene ends, when in the cutscene he was already in the room even further than my character.