All the above.
I fully understand due to circumstances remote is an only option for some. So throw down a bare bones VTT (map and counters) and have at it.
But Roll20, Fantasy Grounds and Foundry all sell themselves as more than that.
Yet:
All that trouble for 20 year old visuals.
Again - no one is expecting AAA here.
But they are not even up to 2013's: The Legend of Zelda: A Link Between Worlds
I think most would be happy with the 2D perspective w/3D polygonal graphics that gives. With little animations for spells, attacks, movement...
Yes basically a video game developer would have to do it. But it's not some out of reach target.
But the point is that it's not simply a linear ramp of improvement from 'bad graphics' through to 'real world graphics'.
1. There are real costs associated with creating a system that can deliver better graphics. The higher fidelity, the more storage and bandwidth is needed.
2. Likewise the clients have to be more capable. If your fan is spinning now, running Foundry and Discord, as it does for many, it will have some serious issues doing much more than that.
3. Production costs are not linear here. One thing is setting up a pipeline to create better assets for a company and delivering those (both to the store, but also into the games, where download speeds are an issue), depending on the level wanted, this can be real expensive.
4. This will have a big impact on creators, who up until now are mainly working in 2D formats; that is images or at most animated maps; universally accepted formats. You make a batch of tokens in PNG, you can deliver them to every platform out there today. But you switch to a 3D format, it's a very different ballgame. This can be eased a bit with cutout standees, flat tokens, etc, but if you want to deliver at that standard, now you're talking about certain newer file formats (e.g. .gltf) and scales (and rotations!) of models and all sorts of stuff, which probably requires tooling (simple though it may be to some extent) on the platform end to manage. Etc.
5. As a GM, you're now in a world where it's a lot harder to match up assets for use in your game, but okay. Moreover though, it's not just screenshotting your PDF and dropping some tokens on there, you're building a 3d world. And that's really time consuming. Maybe that's what you want, and that's okay, but I don't know if most people have tried this (I built 3D levels for a living for a decade). Luckily you could choose your level of engagement here, which seems like a good choice, although I'd personally argue there are better ways to approach this whole thing than this route (see 7.)
6. Players + tokens + map. Easy peasy. You start making that more complicated, you're making the learning curve steeper and the barrier to entry higher. It's hard to get just right if you start adding one more dimension or e.g. a free camera. Some of us are very comfortable with it, others not so much.
But
most important.
7. The bang to buck ratio here is just not worth it by a long shot. If you want a CRPG-like experience with your friends, that's perfectly valid (I haven't used it, but One More Multiverse looks to be doing this), but in many ways heightening the level of abstraction is going to work against a smoother gaming experience. And I don't mean that in a religious OSR-esque theater of the mind sort of way (I can swing that way and most others; I love maps, love props, world creation, etc

), but that the constraints the higher graphics quality bring, beyond a fairly low point, are adversely going to impact the GMs ability to easily prep, stay nimble and not fight the tech and world representation, and likewise for the players it will adversely impact their acceptance of the 'fuzziness' of the shared space and potentially cause barriers to entry (and it's hard enough to get people into Rolemaster Standard System as it is...

).
My $0.02 anyway.