Hook to hv.Table do not work with hv.DynamicMap

ALL software version info

Unix - “Ubuntu 18.04.4 LTS (Bionic Beaver)”
Python Libraries: holoviews 1.12.7, pandas 1.0.1
Browser: Google Chrome

Description of expected behavior and the observed behavior

I present here a simplified example of my original plot to present the specific problem. I have a hv.Table which updates using a hv.DynamicMap (a value selector). My table has some columns which contain long strings, thus I want to customize the width of some of the columns. This is possible using holoviews hooks, as seen in the example below. However, when the hv.DynamicMap updates the table, the column width customization does not take effect, as shown in the screenshots below.

Complete, minimal, self-contained example code that reproduces the issue

%load_ext watermark
import holoviews as hv
import pandas as pd
print('Watermark of notebook execution:')
%watermark -iv

def table(data_id):
    def column_width(plot, element):
        plot.handles['table'].columns[0].width = 1400
        plot.handles['table'].columns[2].width = 1400

    df = pd.DataFrame({
        'x': [data_id*10**12]*4,
        'y': [1, 2, 3, 4],
        'z': [data_id*10**12]*4})

    hv_table = hv.Table(df).opts(
        width=400, height=200, hooks=[column_width])
    return hv_table

dmap_table = hv.DynamicMap(table, kdims=['data_id']).redim.values(
    data_id=[0, 1])

Screenshots or screencasts of the bug in action

The initial plot followed by the plot after selecting another kdims value.

By mistake, I created this issue on Github: https://github.com/holoviz/holoviews/issues/4343.

By mistake, I created this issue on Github: https://github.com/holoviz/holoviews/issues/4343 1.

No mistake, this should really be an issue, thanks for posting it!