@mwaskom is this ok for the original issue?

Yep that resolves the ambiguity of the original text. Should the note also flag matplotlib.colormaps? I guess I’m still slightly confused about what the situation will be in 3.6; will there just be 3 different ways to get a copy of a colormap object? (mpl.colormaps, mpl.cm.get_cmap, mpl.cm.<name>)?

mpl.colormaps will eventually become the default and other cm access methods will be deprecated. However, as long as mpl.colormaps is experimental, I don't want to actively drive people away from mpl.cm.get_cmap. See https://matplotlib.org/stable/users/prev_whats_new/whats_new_3.5.0.html#colormap-registry-experimental for more information.

When the experimental status gets removed, we'll advertise colormaps more loudly.