Streamlit in snowflake with snowpark connector write_pandas deletes columns from table


I see a very strange behaviour when using Snowflake Streamlit with the Snowpark connector:
having this:
inside the form st.experimental_data_editor (because ST 1.22 in Snowflake)
I populate the data_editor with a pandas df which only contains the column I want to show in the data_editor.
num_rows = dynamic

When I insert a new row, fill in all values and then click on the submit button the form is calling:
Session.write_pandas(df, overwrite=True, quote_identifiers=False)
It writes all df data into the table.
All columns which are not in the df get deleted from the table.
Even if there are already rows in that table,
the not used columns in the df get deleted.

I cannot find any information about this behaviour.
This means I will end up with writing all SQLs manually and not using the write_pandas function.