I guess everybody is using the extra bands, please go ahead. It is suggested in the starter notebook:

..Now you have band values associated to each classified pixel of the trining dataset and you can start developing the prediction model. This script extracts only 4 bands, but Sentinel-2 has more: https://developers.google.com/earth-engine/datasets/catalog/COPERNICUS_S2_SR_HARMONIZED#bands.

Am also sure most of us are using similar kind of explainable variables suggested in the starter notebook:

Spectral indexes, derived from band algebra, are also broadly used as explaining variables. Here some example on how to calculate Normalized Difference Vegetation Index (NDVI) and Enhanced Vegetation Index (EVI) over Lansat images: https://developers.google.com/earth-engine/guides/image_math#colab-python

May the best model win!