PSS for CDR circuits

I've tried a bunch of things I've read both on this forum and on cadence boards, but to no avail. I am trying to run PSS + pnoise for the clock output of the clock-data recovery circuit. However, the output load of the VCO has a rather nonlinear input impedance that is dependent both on the VCO swing and the data swing, causing the VCO amplitude to fluctuate a bit, even after reaching steady state, which prevents PSS from converging.

I'm currently using Spectre's PSS and PNOISE. Shooting method with 101 harmonics, and I have verified in the saved tstab waveform that it has reached steady state (I let it run for 10x what I see to be the VCO waveform reaching steady-state, just in case). reltol = 1e-5, vabstol = 1e-7, iabstol = 1e-13. I cannot seem to even run the simulation using Harmonic Balance method.

Any further tips would be appreciated.

Perhaps a rack of transient sims, and tnoise, with whatever
you're looking for taken by postprocessing, is a more
appropriate approach?

After all, a clock recovery from data when the data is
periodically repeating is pretty far from interesting or
from a worst case challenge.