Zig-zag / snap-back behavior in Static Riks capacity curves of transmission tower

Dear Abaqus Community,

I am performing nonlinear collapse analysis of a lattice transmission tower using Abaqus/Standard – Static, Riks to obtain capacity curves (wind speed vs top displacement).
I ran 100 stochastic models (LHS-based uncertainty in material and geometry).

For most models, the capacity curve shows a smooth ascending branch, a peak, and post-peak softening.
However, in some models, I observe a zig-zag / vertical clustering near the peak load:

  • Displacement remains almost constant
  • Wind speed (LPF) oscillates up and down
  • The analysis continues to a descending branch

A representative example is attached.

I suspect this is related to snap-back instability caused by local member buckling and load redistribution, which the Riks method captures through non-monotonic equilibrium paths.

My questions

  1. Is this behavior physically meaningful (progressive/local instability), or does it indicate a numerical issue?
  2. Is such zig-zag behavior expected in Static, Riks for lattice structures experiencing snap-back?
  3. For fragility/capacity assessment, is it acceptable to use the upper envelope or peak capacity, while acknowledging this behavior?