> I suppose that to some degree [*] the main benefit of this patch is that
> it offers an option that calls quit-restore-window.

Yes, I agree with this.  Setting org-src-window-setup to other-window
was almost good enough for me -- it even respected
display-buffer-base-action -- except that it wouldn't close the popped
up window.

> And that makes me think that the current options that go through a
> simple display-buffer-based call (current-window and other-window) would
> benefit from calling quit-restore-window like your `plain` option does.
> If you agree, perhaps it's worth adding another patch on top that does
> that.

I agree other-window would benefit from quit-restore-window, it makes
sense to close the existing window if it's been popped up.

I'm less sure about current-window.  We could certainly call
quit-restore-window here, but I'm not sure there's any benefit, as it
shouldn't open new windows that need to be closed (unless you're
modifying display-buffer-alist, in which case the `plain' option should
be preferred anyways).  I'm also hesitant because Samuel relies on this
option for accessibility reasons, and if I accidentally introduced a
bug, I might not immediately notice, since I don't use this option.

I've attached a patch on top of my previous one, calling
quit-restore-window when using other-window, but leaving current-window
alone.

> Hmm, weird.  I tried again (Emacs 26.3, vanilla config) and still see
> the behavior I reported.  Oh well.

I tested again, with "emacs -q" this time, and got the behavior you
reported. So it must be something with my config.