Actually, triangulation is pretty much always the first step for new areas - the terrain usually to be triangulated, too! We start with a totally blank slate, and block in as much geometry as we can, in the correct spots and at the correct size and rotation, before we start high-detail modeling.
In some cases, detail models can be made before triangulation is finished. This happens a lot with manmade objects - they are generally pretty geometric in nature, so it's not really necessary to triangulate (or "camera match") to determine their shape, only their placement in the scene.
For most of our assets, the modeling process is generally: camera matched geometry -> high-poly model/(or sculpt in some cases) -> uv unwrapping -> texturing -> low-poly model. As soon as any
version of the asset exists, we bring it into the engine so we can start working on interactivity and material settings, plus just start getting a feel for the area in realtime. So that part is kind of always happening, as models are tweaked or textured and get re-imported to the engine along the way.
We do try to split up the game into manageable chunks while working (chunks like Prison Island, the Survey Island elevator room, the Jungle Island village basin, and 233...), and then those chunks are divided up into individual assets. Those assets are then split into various tasks (like modeling vs texturing vs animation). The next part is fun: we use a service called Trello to oranize all of this info into to-do lists, which ends up looking like this:
This allows us to parcel out the mountain of content that is Riven into individual pieces that an artist can tackle, one at a time. Someday, they will all be finished.
I'm not really the best to comment on all of the nitty-gritty of what I just described, but that is (I think) a pretty decent explanation of how we work, in general.