Sorting lost in Altair layered bar chart with error bars

I am using a custom sort for my bar chart and it works well. However when I want to add error bars to it and use a layered chart then the sorting is not taking into account anymore. I also defined axis = None and that is also not taking into account.

Here is an example of the data:

df = pd.DataFrame(
    {'size' : ['huge', 'huge', 'huge', 'huge', 'huge', 'huge', 'big', 'big', 'big',  'big', 'big', 'big', 'small', 'small', 'small', 'small', 'small', 'small'],
     'weight': ['10 mg', '10 mg', '10 g', '10 g', '10 kg', '10 kg', '10 mg', '10 mg', '10 g', '10 g', '10 kg', '10 kg','10 mg', '10 mg', '10 g', '10 g', '10 kg', '10 kg'],
     'value': [3.5,2.6,5.1,6.5,2.3,4.6,7.1,2.8,6.9,1.5,2.6,2.8,6.9,2.3,4.6,3.5,2.6,5.1]
    }
)

Using just the bar chart works

alt.Chart(df).mark_bar().encode(
    x = alt.X('weight:O', title=None, axis=None, sort=['10 kg', '10 g', '10 mg']),
    y = alt.Y('mean(value)', title='Value'),
    color = alt.Color('weight:O', sort=['10 kg', '10 g', '10 mg']),
    column = alt.Column('size', sort=['huge', 'big', 'small'])
)

But not anymore with the error bars:

error_bars = alt.Chart().mark_errorbar(extent='ci').encode(
    x=alt.X('weight:O', sort=['10 kg', '10 g', '10 mg']),
    y='value:Q'
)

bars = alt.Chart().mark_bar().encode(
    x = alt.X('weight:O', title=None, axis=None, sort=['10 kg', '10 g', '10 mg']),
    y = alt.Y('mean(value)', title='Value'),
    color = alt.Color('weight:O', sort=['10 kg', '10 g', '10 mg'])
)

alt.layer(bars, error_bars, data=df).facet(
    column = alt.Column('size', sort=['huge', 'big', 'small'])
)

In both plots the axis and title have been set to None but is not taking into account in the layered chart. The weird thing is that the sorting is taking into account for the legend (see color = ...) but not for the x-axis (within each size).

Is there a way around this or am I not using the layered charts correctly?