Dear Matteo,
At certain yaw errors in the idling condition, a known blade-edgewise instability exists, e.g., as discussed in the following forum topic: Designing for yaw errors using FAST - #7 by Jason.Jonkman. When the model is physically unstable, a steady-state solution cannot be reached without adding damping damping to the system during the steady-state solve (e.g., via Bld_Kdmp) or following the other suggestions in the linked forum topic. Alternatively, you could linearize at time zero (before the unstable solution results in excessive motion) to quantify the eigensolution approximately (I say, “approximately”, because the linear model is not fully representative of the underlying dynamics if the system is not in steady state before linearizing).
Yes, the MBC procedure and Campbell toolbox can work for cases where you only output the linearization at one azimuth angle because the rotor speed is zero.
Best regards,