If we can be pretty confident it'll be 100 years, that's long enough for technology to progress a lot before we fiddle with it. Not sure what "soon" means in this case, but I suspect they don't mean "within a couple years".
Like there being infinite number of infinities that are ever larger, there's a seemingly infinite number of soons, each further away in time than the previous.
So I propose we invent something like aleph[1], but for soon. This way we can communicate clearly just how soon soon is.
Technological advancement doesn't happen Inna vacuum. To get good at things you have to try them. Getting good at the sort of technology required to build such a project requires trying such a project. It's not a matter of waiting 100 years, it's a matter of trial and error for 100 years.
This is true, but large engineering projects aren't a single technology. There are a lot of ways we'll improve relevant technologies without actually working on this directly. Presumably the fracking industry is improving our ability to model the subsurface, for example.
This also means other industries (such as fracking) would learn a lot from a massive Yellowstone geoengineering project, but other projects and smaller geothermal projects should make this one more feasible and predictable.
(This is how I feel about geoengineering in general: it's expensive and risky, but may end up being necessary, so let's practice on a smaller, safer scale before massive, dangerous projects become urgently needed.)
Not sure what kind of scientist could say they don't expect it to ever erupt again?