Better explanation of 'minimal' in xarray.open_mfdataset(data_vars='minimal') in docs?

Problem description

I'm currently troubleshooting some overly long (to me) load times using open_mfdataset on GFS data. In trying to speed things up, I'm trying to specify just the four variables I actually care about using data_vars=[strings], but to no avail. It still takes ~30 minutes to load 52 time slices from 7 files.

In the docs I do see that if data_vars =

list of str: "The listed data variables will be concatenated, in addition to the ‘minimal’ data variables."

However, I can't seem to understand what the 'minimal' variables are from this sentence in the docs:

‘minimal’: Only data variables in which the dimension already appears are included.

All the variables in the CF-compliant GFS data are associated with dimensions. So does that mean that all the variables in the files will be concatenated, regardless if I specify which ones I want? I feel like I'm misunderstanding what is included by default.

Start with the open_mfdataset API documentation linked in the issue and trace the meaning of data_vars='minimal' and its interaction with an explicit variable list. Update the explanation so readers can identify which variables are included by default and verify the wording against the reported GFS use case.