I want to get a multiscene layout described in https://docs.enthought.com/mayavi/mayavi/auto/example_multiple_mlab_scene_models.html

import numpy as np

from traits.api import HasTraits, Instance, Button, \
    on_trait_change
from traitsui.api import View, Item, HSplit, Group

from mayavi import mlab
from mayavi.core.ui.api import MlabSceneModel, SceneEditor


class MyDialog(HasTraits):

    scene1 = Instance(MlabSceneModel, ())
    scene2 = Instance(MlabSceneModel, ())

    button1 = Button('Redraw')
    button2 = Button('Redraw')

    @on_trait_change('button1')
    def redraw_scene1(self):
        self.redraw_scene(self.scene1)

    @on_trait_change('button2')
    def redraw_scene2(self):
        self.redraw_scene(self.scene2)

    def redraw_scene(self, scene):
        # Notice how each mlab call points explicitly to the figure it
        # applies to.
        mlab.clf(figure=scene.mayavi_scene)
        x, y, z, s = np.random.random((4, 100))
        mlab.points3d(x, y, z, s, figure=scene.mayavi_scene)

    # The layout of the dialog created
    view = View(HSplit(
                  Group(
                       Item('scene1',
                            editor=SceneEditor(), height=250,
                            width=300),
                       'button1',
                       show_labels=False,
                  ),
                  Group(
                       Item('scene2',
                            editor=SceneEditor(), height=250,
                            width=300, show_label=False),
                       'button2',
                       show_labels=False,
                  ),
                ),
                resizable=True,
                )


m = MyDialog()
m.configure_traits()

Each scene has to render a separate volume object.

I have provided a custom redraw_scene function with

def redraw_scene(self, scene):
        # Notice how each mlab call points explicitly to the figure it
        # applies to.
        mlab.clf(figure=scene.mayavi_scene)
        s = np.random.random((100, 100, 100))
        mlab.pipeline.volume(mlab.pipeline.scalar_field(s), figure=scene.mayavi_scene)

but ended up getting both volumes rendered on the second scene.

I have also tried the setup with a separate engine per scene but it yields the same result.

How I do get volume renders in separate scenes with Mayavi?