Switching back and forth between dcc.Tabs doesn't seem to release memory in the browser

In a dash app with a 50 000 point scatter chart in tab 1 and another in tab 2, switching back and forth between those tabs increases the memory footprint that I see in the Chrome task manager by about 100 MB every time but it doesn't look like any gets released.

In the snapshot above, the memory footprint has climbed to 1.5 GB. Within a few minutes of not using the app it dropped, but only to 1 GB so it's still way higher than it was before I'd interacted with it.

Also, it looks like others in the community have seen something similar.

Describe your context
Please provide us your environment, so we can easily reproduce the issue.

• replace the result of pip list | grep dash below

dash                            2.17.0
dash-ag-grid                    2.1.0
dash-bio                        1.0.2
dash-bootstrap-components       1.4.2
dash-bootstrap-templates        1.1.2
dash-chart-editor               0.0.1a4
dash-core-components            2.0.0
dash-cytoscape                  0.3.0
dash-dangerously-set-inner-html 0.0.2
dash-design-kit                 1.10.0
dash-embedded                   2.14.0
dash-enterprise                 1.0.0
dash-enterprise-auth            0.1.1
dash-enterprise-libraries       1.4.1
dash-extensions                 0.1.6
dash-facebook-login             0.0.2
dash-gif-component              1.1.0
dash-html-components            2.0.0
dash-iconify                    0.1.2
dash-mantine-components         0.12.1
dash-notes                      0.0.3
dash-renderer                   1.9.0
dash-snapshots                  2.2.7
dash-table                      5.0.0
dash-user-analytics             0.0.2

• if frontend related, tell us your Browser, Version and OS

  • OS: Unbuntu 22.04
  • Browser Chrome
  • Version 122

Example app:

import dash
from dash import dash_table
import dash_core_components as dcc
import dash_html_components as html
from dash.dependencies import Input, Output
import plotly.graph_objs as go
import numpy as np
import pandas as pd

# Data generation for plots and table
np.random.seed(42)  # for reproducibility
x = np.linspace(0, 10, 50000)
y1 = np.random.randn(50000)
y2 = np.sin(x) + np.random.randn(50000) * 0.2

table_data = pd.DataFrame(np.random.randn(10, 4), columns=list("ABCD"))

# App initialization
app = dash.Dash(__name__)
server = app.server

# Layout
app.layout = html.Div([
    dcc.Tabs(id="tabs", value='tab-1', children=[
        dcc.Tab(label='Scatter Plot 1', value='tab-1'),
        dcc.Tab(label='Scatter Plot 2', value='tab-2'),
        dcc.Tab(label='Data Table', value='tab-3'),
    ]),
    html.Div(id='tabs-content')
])

# Callback to update tab content
@app.callback(Output('tabs-content', 'children'),
              Input('tabs', 'value'))
def render_content(tab):
    if tab == 'tab-1':
        return dcc.Graph(
            id='scatter-plot-1',
            figure={
                'data': [go.Scatter(x=x, y=y1, mode='markers')],
                'layout': go.Layout(title='Scatter Plot 1')
            }
        )
    elif tab == 'tab-2':
        return dcc.Graph(
            id='scatter-plot-2',
            figure={
                'data': [go.Scatter(x=x, y=y2, mode='markers')],
                'layout': go.Layout(title='Scatter Plot 2')
            }
        )
    elif tab == 'tab-3':
        return html.Div([
            html.H4('Data Table'),
            dash_table.DataTable(
                data=table_data.to_dict('records'),
                columns=[{'name': i, 'id': i} for i in table_data.columns]
            )
        ])

if __name__ == '__main__':
    app.run_server(debug=False)

Run the supplied Python example app and reproduce the memory increase by switching between dcc.Tabs values. Start at the tabs-content callback and its render_content entry point, then use Chrome task manager while switching tabs repeatedly. Done means repeated tab switching no longer leaves the browser memory footprint continually elevated.