Hm... I still can't get something. So I linearise it as ##f = 2kl(1 - \frac{l}{s})\sin\theta##, then I project the weight on the same axis and say that total force is ##f + W_{projection}##, which isn't really a linear combination of ##Δx##, so not a restoring force?
If I take ##\sum{\tau}=\tau...