To my mind, for any data set to be of any kind of use you would want to fire hundreds of rockets around the storm simultaneously, and repeatedly, in order to sample every piece of air going into and coming out of the storm. Otherwise, you get one trajectory of data and that's it - how will that help forecasting? Models are still pretty/very poor at forecasting exactly where convection will fire - if you want to be able to warn people quite far in advance, you need to be able to resolve the mesoscale reasons for storm formation, and get them to be simulated in models.
Of course, we still want to know the fundamental reason for tornadogenesis, but even if/when we do find it out, unless we can back that up with models which can cause thunderstorms to develop in the correct locations, and simulate the inflow parcels, and exact microphysics of that particular storm cell, I don't see how lives can be saved by such research.