WRF v3.9.1: d02 has mostly zero SFX output while d01 and d03 are normal

[aurorrmok]

I’m encountering a strange issue with WRF v3.9.1 and would appreciate any insights.
I’m running WRF with ERA5 initial/boundary conditions, using three nested domains (d01–d03).

• d03 is a vortex-following moving nest
• The simulation completes successfully without crashes

However, when analyzing the SFX variable in wrfout files, I find that:

• d01 and d03 have complete, continuous SFX output
• d02 only has nonzero SFX values at a few time steps, while most times are exactly zero

Other surface variables (e.g., U10/V10, LH, TSK) appear normal, so this doesn’t seem to be a general surface physics failure.
Has anyone seen domain-dependent zero SFX output before, especially in simulations with a vortex-following nest?
Any suggestions on where to look (namelist, physics, I/O, or known WRF quirks) would be greatly appreciated.