Now don't get me wrong, yes I know it's Rockstar, yes they are probably still going to make millions because GTA has such an established fanbase and they will buy a literal turd with a rockstar logo on it given the opportunity, this is all true.
However even if the story mode is half-decent and I've got this feeling it's going to be fairly lackluster given the angle that they're going with and it seems despite being a 'crime' game they're shying more and more away from controversy these days and have definitely been infested with the woke.
Here's where it's going to get funny though, GTA Online. Oh boy is that going to be a clusterfuck. The brain drain is very real at game studios and it looks like they're having more and more problems holding onto talent. Any programmer with half a brain isn't going to want anything to do with these arseholes. Found a youtube video that does a perfect job encapsulating all the problems with online which as with most modern games is just a vehicle for micro-transactions so they make sure it's always more convenient to whip out your bank card and sign over your details. It's remarkable how Rockstar Games have managed to make a crime game that's boring.
https://www.youtube.com/watch?v=5IypemNHafM
I have barely even ranted about how bad this game is going to be from a technical networking perspective yet either since we're seeing how GTA V even to this day is constantly falling over due to the number of players. Even if people still buy the game, they're going to get pissed off at the state of it. I have to wonder as well how much of the back end of GTA VI they've really messed with and if it's just another clone of the same code as GTA V.
I'd seen BeamNG used spring-mass systems for their deformation, but given that damage was limited to certain zones in GTA V, I'd assumed they used pre-authored damage morphs & debris meshes like everyone else. Got any links, it would be an interesting read?
As for the expense, requirements dependent, you might be surprised. I played around with soft body in Unreal a bit. A really performant PBD implementation was actually quite quick to get up and running. Non-linear Gauss-Seidel solvers can be heavily parallelized on GPU through clever clustering and graph colouring. Collision remains a bit of a sticking point - I got about half-way towards good results with low resolution proxy hulls, but then Unreal went and changed physics APIs.