Matplotlib figures are not redrawn in the cell in which they are created

Describe the issue

Using an interactive figure's canvas.draw() method does not update the figure if it was created in the same cell as the draw call.

Figure created in the same cell:

Figure created in another cell:

Reproduce

1. Open a new notebook in JupyterLab or Notebook
2. Insert the following code:

import matplotlib.pyplot as plt
import time

%matplotlib widget

fig = plt.figure()

for i in range(5):
    x = list(range(i+2))
    xx = [x**2 for x in x]
    plt.clf()
    plt.plot(x, xx)
    fig.canvas.draw()
    
    time.sleep(1)

3. Run the cell and observe that the plot does not show until the loop has exited.

Optional:

1. Move the for loop to a separate cell from the figure creation
2. Run cell and observe plot updates during the loop

This can be reproduced using the scipy notebook docker stack.

Versions

3.8.5 | packaged by conda-forge | (default, Jul 31 2020, 02:39:48)
[GCC 7.5.0]
ipympl version: 0.5.7
jupyter core     : 4.6.3
jupyter-notebook : 6.0.3
qtconsole        : not installed
ipython          : 7.17.0
ipykernel        : 5.3.4
jupyter client   : 6.1.6
jupyter lab      : 2.1.5
nbconvert        : 5.6.1
ipywidgets       : 7.5.1
nbformat         : 5.0.7
traitlets        : 4.3.3
Known nbextensions:
  config dir: /opt/conda/etc/jupyter/nbconfig
    notebook section
      jupyter-matplotlib/extension  enabled
      - Validating: OK
      jupyter-js-widgets/extension  enabled
      - Validating: OK
JupyterLab v2.1.5
Known labextensions:
   app dir: /opt/conda/share/jupyter/lab
        @bokeh/jupyter_bokeh v2.0.3  enabled  OK
        @jupyter-widgets/jupyterlab-manager v2.0.0  enabled  OK
        jupyter-matplotlib v0.7.3  enabled  OK

Comienza ejecutando la reproducción proporcionada en la misma celda en JupyterLab o Notebook con el backend %matplotlib widget; después, compárala con la versión en celdas separadas. Se considera terminado cuando fig.canvas.draw() actualiza visiblemente la figura durante cada iteración del bucle, en lugar de hacerlo solo después de que termine el bucle.