No this is not a bug. This is completely intended!!! We cannot call any of the biology models from initial because we will introduce a big adjoint bug!!!
The way that I solved this problem is to rewrite the correct value for the attribute bio_file in the header of output NetCDF file, when the file is closed. Check routine netcdf_close in file mod_netcdf.F. The fact that you are having this problem implies that you are either checking the file before it is closed or you are not using the call to:
CALL netcdf_close (ng, model, ncid, ncname)
everytime that a NetCDF file is closed. This is part of the new ROMS I/O design. This routine is generic and expandable, so in the future we can take care of things like this.
Anyway, I think that I forgot to document this so I don't expect anybody to know about this trick.
I had to do this trick because now we are allowing to have multitple copies of the biological models. For example, you can have a modified copy of any of the biology models in the working (application) directory. John Wilkin asked for this capability since usually the biological models are tuned (changed a little) for a particular application. This will avoid to making changes to the distribuited version of these models.