Okay, I'm on it

This is my report for pytest mne/viz/_brain on mne-python/ref:

1. pyvista/master, pyvistaqt/master:
   all tests are green

2. pyvista/mem, pyvistaqt/master:
   Segmentation fault in test_brain_init
   Does not crash if I use -k init and the test is green

3. pyvista/master, pyvistaqt/mem:
   MemoryError in test_brain_traces for volume and mixed, test_brain_linkviewer

4. pyvista/mem, pyvistaqt/mem:
   Segmentation fault in test_brain_traces
   Does not crash if I use -k traces and I got the same MemoryError as in 3)

I think the MemoryError is related to your traceback, I'll investigate

I identified the issue with MemoryError on 3). BasePlotter.deep_clean() is forced in pyvistaqt/mem

deep_clean() will set self.parent = None which is used to remove the actors in our function clear_points().

I would recommend to proceed with caution here. PyVista made the difference between 'close' and 'clean'. I suggest that we stick to it until we can have a proper discussion or plan in pyvista/pyvista#959

The way I see it, Brain -> BackgroundPlotter -> BasePlotter (-> : depends on) so the sequence should be (in this order):

Brain.close(), BackgroundPlotter.close(), BasePlotter.close(), Brain.clean(), BackgroundPlotter.clean(), BasePlotter.clean()

I identified the issue with the segfault on 2). On our code, the patch to remove lights should be removed and self.plotter.interactor=None needs to removed. The refs to vtkXOpenGLRenderWindow will remain though.

EDIT: I can work something out, it's just some cleaning to do

Brain.close(), BackgroundPlotter.close(), BasePlotter.close(), Brain.clean(), BackgroundPlotter.clean(), BasePlotter.clean()

Do you mean BasePlotter.deep_clean?

Do you mean BasePlotter.deep_clean?

Yes. I used 'clean' just to give the idea

I reverted the deep_clean in pyvistaqt and now things pass there and on this PR!

I'll go ahead and merge this since it has a fix for latest matplotlib