PySpark - Combine calculation and kwargs in 'select' in a 'for' loop

I was told using withColumn within a for loop is not a good idea in terms of performance and I try to replace it with select, as suggested. But I don't think it's a good idea to write all columns (and yes, I'm a bit lazy^^); I already have a dict with the list of columns and I try to use it, but I get an error at the second loop.
My main issue is that I need at least one column for the calculation, but **kwargs argument come after the others, so I try to make it dynamic: remove the column from the list, pass the column first, then my calculation, finally the other columns (of course, the loop is over the column list).
The code runs in a palantir-foundry code repository (I have no other pySpark environment to test).

Here is a snippet of my code, I hope it's enough to make it reproductible:

input_schema = {
    'col_1': T.StringType(),
    'col_2': T.StringType(),
    'col_3': T.StringType(),
    'col_4': T.StringType(),
    'col_4': T.StringType(),
}


@tranform_df(
    Output("output_df"),
    nsw=Input("My_Input")
)
def my_compute_function(nsw):
    nsw = nsw \
        .withColumn("errors",  F.lit(None).cast("array<string>")) \
        .withColumn("error_check",  F.lit("Pouet").cast("string"))

    for col_name, test_list in validation_list.items():

        l_input_schema = dict(input_schema)
        del l_input_schema[col_name]

        for to_test in test_list:
            nsw = nsw \
                .select(
                    *col_name,
                    F.when(
                        (nsw.error_check.isNotNull()),
                        F.when(nsw.errors.isNull(), F.array(nsw.error_check))
                        .otherwise(F.array_union(nsw.errors, F.array(nsw.error_check))),
                        )
                    .otherwise(nsw.errors)
                    .alias("errors"),
                    **l_input_schema
                )

    return nsw

No syntax error here, but at the second loop, the following message:

Function select() got an unexpected argument col_2. Please review your code.

The error points out the line where kwargs is used (**l_input_schema), and it's always the second column