If you don't find an answer here, there may be some guidance on the High Resolution Discussion Board, where you can also ask a question: https://bb.cgd.ucar.edu/cesm/forums/high-resolution-variable-resolution.163/
(1) Vertical level issue when generating initial conditions
When using NCL codes to create initial conditions from a previous CAM run, it was discovered that the model top pressure altitude of the CAM run must be higher at all locations than the specified vertical levels when generating "real" file from the MPAS-Model directory. This issue arose when attempting to use a 32L CAM-Chem run with a 32L MPAS model grid with a top above 40 km. In the winter hemisphere, the top of the 32L CAM-Chem simulation was lower than 40 km, and entire columns for the dynamical field (e.g., rho, theta) were getting set to the fill_value (1e36) which eventually prevented MUSICAv1 from running.
The recommendation from Bill Skamarock (MMM) was to use a simulation with a sufficiently high model top (e.g., 70L WACCM) to regrid for initial conditions so this issue does not arise. This is thought to be preferable than reducing the model top of the MPAS model, which could have unintended consequences on the eventual MUSICAv1 simulation.
It is expected that there will be a forthcoming replacement to these NCL scripts with Python which will remove this issue.