I would say these are all valid methods, what we’re looking for is speed and good end results. Yet I think what sets them apart is the optimization. You could do the tire pattern with multires and alpha brushes, sculpt it, etc… but the resulting vertex count should be insanely dense, then having to remesh, you loose control over your topology, not very controllable is not that cool if you ask me, not easy to make changes and redesigns and a lot of wasted geometry and memory, etc, etc. So at least I wouldn’t go about modeling a tire with sculpt mode in mind. Not to say it isn’t valid, of course you can.
I think it is best to approach the modeling with good old box modeling. It would definitely result in the most optimized and controllable mesh, and for me that trumps most of the approaches that tend to want to make the job very easy and quick. As Jonathan Lampel says, it can be tedious, but for me, that I can have mastery and authority over me loops, it’s worth it.
When it comes to geometry nodes, I wouldn’t know, I’ve never used them. Though I doubt right now it would be better than modeling the tire. I believe they are most useful for motion graphics and procedural patterns. You could create the alpha with geometry nodes, but again, I find it too much of a hassle to go through geometry nodes just for that. Creating the alpha in an application like Photoshop or Krita or Illustrator should be the best way.
