Why would a callback return results from former calls without caching?

Dash Python
1

I have this callback that created different kinds of plots and it worked as expected. All of a sudden this callback returns results from past calls. E.g. I create the histograms but the boxplots are still returned. I am puzzled. There is no caching, and the arguments are passed correctly. Why would this callback return boxplots if boxplot is not provided in kinds?

@app.callback(

Output('qc-figures', 'children'),

Input('qc-update', 'n_clicks'),

State('tab', 'value'),

State('qc-groupby', 'value'),

State('qc-graphs', 'value'),

State('qc-select', 'value'),

State('wdir', 'children')

)

def qc_figures(n_clicks, tab, groupby, kinds, options, wdir):

    if n_clicks is None:

        raise PreventUpdate

    print('groupby:', groupby)

    print('kinds:', kinds)

    print('options:', options)

    df = get_complete_results( wdir )

    sns.set_context('paper')

    

    by_col = 'Label'

    by_col = 'Batch'

    quant_col = 'peak_max'

    quant_col = 'log(peak_max+1)'

    if options is None:

        options = []

    figures = []

    n_total = len(df.peak_label.drop_duplicates())

    for i, (peak_label, grp) in tqdm( enumerate(df.groupby('peak_label')), total=n_total ):

        if not 'Dense' in options: figures.append(dcc.Markdown(f'#### `{peak_label}`', style={'float': 'center'}))

        fsc.set('progress', int(100*(i+1)/n_total))

        df = get_complete_results( wdir )

        if len(grp) < 1:

            continue

        if 'hist' in kinds: 

            fig = sns.displot(data=grp, x=quant_col, height=3, hue=groupby, aspect=1)

            plt.title(peak_label)

            src = fig_to_src(dpi=150)

            figures.append( html.Img(id='hist', src=src, style={'width': '300px'}) )

        if 'boxplot' in kinds: 

            fig = sns.catplot(data=grp, y=quant_col, x=groupby, height=3, kind='box', aspect=1.3, color = "w")

            if quant_col in ['peak_max', 'peak_area']:

                plt.gca().ticklabel_format(axis='y', style='sci', scilimits=(0,0))

            plt.title(peak_label)

            src = fig_to_src(dpi=150)

            figures.append( html.Img(id='box', src=src, style={'width': '300px'}) )

        if not 'Dense' in options: figures.append(dcc.Markdown('---'))

        if i == 3: break

       

    return figures


def fig_to_src(dpi=100, format='png'):
    out_img = io.BytesIO()   
    plt.savefig(out_img, format=format, bbox_inches='tight', dpi=dpi)
    plt.close('all')
    out_img.seek(0)  # rewind file
    encoded = base64.b64encode(out_img.read()).decode("ascii").replace("\n", "")
    return "data:image/png;base64,{}".format(encoded)

When I look at the generated figures, there are as many figures produced as expected. It seems the output container is not cleared correctly.

I think the problem is with multiple images the id= were not unique. I removed the id argument and now it is working as expected. But why was it working before and switched to the unexpected behavior out of the blue?