I am not entirely certain if I am correct, but in discussion about the RROPAMP in below thread, my observation is that the model of the RROPAMP consists of two poles (pole1,pole2) and a zero (zero1). This zero is a right-half plane (RHP) zero, and based on the profile of V(out2), the spike appears to be the effect of RHP zero (i.e inverse response).
If you increase the GBW (gain-bandwidth product) to a very large value, such as 500 MHz, it can effectively move that pair of zero and pole (zero1 and pole2) an orders away from your signal frequency. As a result, the weird spike disappears.
Rail-to-Rail Output OpAmp (Ã-Device RROPAMP) Instance Param - QSPICE - Qorvo Tech Forum
