DateTime formatter issue


I have a dataframe (link) with two datetime columns that I use to calculate the datetime difference (timedelta). I do it as follows:

import pandas as pd

dfs = pd.read_csv('', header=[0], parse_dates=['Data do Download', 'Data da Observação'])
dfs['Diferença de Tempo'] = (dfs['Data do Download'] - dfs['Data da Observação']) - timedelta(hours=3)
dfs['Diferença de Tempo'] = pd.to_timedelta(dfs['Diferença de Tempo'])

The resulting dataframe is:

Then, I display this dataframe within a panel app, and I realized that the datetime difference (from the ‘Diferença de Tempo’ column) shows different values with the pn.pane.DataFrame, pn.widgets.DataFrame, pn.widgets.Tabulator and pn.pane.Perspective widgets.

With the pn.pane.DataFrame, I get the correct values:

With the pn.widgets.DataFrame, I use the the formatters={'Diferença de Tempo': models.DateFormatter(format='%d days %H:%M:%S'),} option to format the datetime difference as it is shown within the pn.pane.DataFrame, and there is a time difference of one day:

With the pn.widgets.Tabulator, using the same formatters option, I get the same result with the one-day time difference:

And with the pn.pane.Perspective, I could not figure the correct way to format the date:

Finally, I would like to stick with the pn.widgets.DataFrame, but I need to figure out how to properly display the datetime difference (in the format %d days %H:%M:%S) in the dataframe. The full code is available here. Maybe I’m not correctly handling the datetime columns in my dataframe? Any answers and guidance on this issue are greatly appreciated!

Thank you,