The following change in ft_text_renderer::update_line_bbox implements what I proposed in comment #1:

diff -r 6bf833d96cbe libinterp/corefcn/ft-text-renderer.cc
--- a/libinterp/corefcn/ft-text-renderer.cc     Sat Feb 11 09:57:47 2023 -0800
+++ b/libinterp/corefcn/ft-text-renderer.cc     Mon Feb 13 20:55:59 2023 +0100
@@ -775,6 +775,11 @@
       // Add one pixel to the bbox height to avoid occasional text clipping.
       // See bug #55328.
       bb(3) = (m_ymax + 1) - m_ymin;
+
+      // Fix a minimum line height of 0.5 fontsize, see bug #57352.
+      if (bb(3) < 0.5 * m_font.get_size ())
+        bb(3) = 0.5 * m_font.get_size ();
+

But it doesn't actually make a visual difference. Changing to 1 fontsize makes it better but still not enough.

@Pantxo: I think this will be quick.  Do you know where in the code to stick a statement so that there is a minimum vertical spacing that scales with "FontSize"?

Okay, I didn' realize that Matlab also produced lines with varying vertical spacing.  I tried this

clf;
text (.5, .5, {"a", "b^C_D", "e", "f", "", "g"})
text (.6, .5, {"a", "bCD", "e", "f", "", "g"})

and got the results attached in file vspace_example.png.

I don't think there is anything we need to do for Octave in that regard.

I think it would be nice to have a blank line produce a vertical space that was closer to what an ordinary character would make.  Re-titling bug report for that issue.

When there are multiple lines in a text() command, Matlab uses the same vertical spacing for every line.  Octave seems to inspect each line and use a different vertical spacing based on the characters present (characters with descenders or CAPITAL letters will have different heights).

Sample code is attached

## '[]' should create a newline with same height as regular line
clf;
text (.3, .8, {"Hello", 'world', [], 123}, 'fontsize', 20)
text (.5, .8, {"Hello", 'world', 1, 123}, 'fontsize', 20)

The resulting plot is attached as well.