This is wonderful. This makes so much more sense now:

import numpy as np

import mne
from mne.datasets import sample

data_path = sample.data_path()

fname = data_path + '/MEG/sample/sample_audvis-ave.fif'

# Reading evoked data
condition = 'Left Auditory'
evoked = mne.read_evokeds(fname, condition=condition, baseline=(None, 0),
                          proj=False)
evoked_meg = evoked.copy().pick_types(meg=True, eeg=False)

evoked_meg.plot_topo(ylim=dict(grad=[-40, 40], mag=[-100, 100]),
                     scalings=dict(grad=1e13, mag=1e15))

Is there an example that we could perhaps update? Does this affect any existing example?

actually I'm not sure if it's fully fixed. Try changing the ylim for mag to something high. It doesn't affect the plot ...

evoked_meg.plot_topo(ylim=dict(grad=[-40, 40], mag=[-2000, 2000]),
                     scalings=dict(grad=1e13, mag=1e15))

Looks like the new bug is from the unzipping process. Will test and try to fix it!

Still needs a bit of testing !

Still something is weird. Compare:

evoked_meg.plot_topo(ylim=dict(grad=[-0.5, 0.5], mag=[-50, 50]),
                     scalings=dict(grad=1e13, mag=1e15))

and

evoked_meg.plot_topo(ylim=dict(grad=[-0.5, 0.5], mag=[-300, 300]),
                     scalings=dict(grad=1e13, mag=1e15))

I didn't touch the gradiometer limits but they changed too ...

Yeah, I just tried it and can confirm that the grad limits change too. Maybe it's because the list of lists is being unzipped somewhere else and that's messing it up?

I guess you can set the limits manually above the for loop we were looking at to see if your hypothesis holds true. Try with both the options above.

Looks like it was a case of mixing the limits up between channel types (within that loop) due to the zipping and unzipping process.

That last commit should deal with previous CI issues !

thx a lot @ramkpari !