It may be depending on how exactly SVP handles the interpolation (which I'm obviously not privy to), and why I asked the question. I would also suggest you treat your customers kindly and patiently, not like a fellow developer who should know every last detail of the software's inner workings. If it "makes no sense" because the results would be exactly the same, (then I agree with you), but an explanation (instead of an short and insulting answer) would be much appreciated. The people here are the ones who are feeding you, so it would be recommended to show an ounce of patience with them. It is just successful business 101.
The reasoning behind my theory is to keep the frames synced up to realtime / integer equivalent images needed for the final 90fps since it isn't a simple integer multiplication. We'd just discard the frames we didn't need on the 180fps (intermediate video) down to 90fps conversion. This would theoretically be advantageous if SVP only did linear motion blending instead of a cubic or higher order motion vector interpolation. It has been stated multiple places that fractional interpolation yields sub optimal results from SVP, and I was wondering if this method would help improve it since we aren't relying on doing this expensive calculation real-time.