If you use the “take it to extremes” method to try to see what’s going on...then visualize a bow with a 90 degree, right angle bend. How far into the draw would it take for the string lift off and only touch at the tips?
To the other extreme...if you made the static bend barely perceptible until your tips reached the same position in front of the back as the first extreme example, the bow would need to be longer than the tree it came from, and if that whole portion were static, then that would cause ridiculous mass beyond the working portion of the limb...and seems like it would only represent a stiff tipped reflex.
Visualizing those extremes, it seems to me that example 1 would come closer to producing the desired effect of recurves vs example 2. So...it seems like the sweet spot would be something toward the “tighter radius” end of the spectrum. Also, it seems like a tighter radius would mean a shorter non-working area of limb...meaning more limb able to work and less tip mass to-boot.
That’s my theoretical take on the question...which I fully understand could be highly flawed based on possible misunderstand of what I’ve read and studied. I also tend to believe real world vs theoretical, and we’ve already had one vote for sweeping curves vs tight curves...so...there’s that.