One of the unit tests in inst/ls_distance_fcn may fail randomly, according to the load of the system on which it is executed:

% Compare timing to ls_signed_distance.
%!test
%!  n = 500;
%!  x = linspace (-10, 10, n);
%!  h = x(2) - x(1);
%!
%!  [XX, YY] = meshgrid (x, x);
%!  phi = ls_genbasic (XX, YY, "sphere", [0, 0], 8);
%!
%!  id = tic ();
%!  d = ls_distance_fcn (phi, h);
%!  time1 = toc (id);
%!
%!  id = tic ();
%!  sd = ls_signed_distance (phi, h);
%!  time2 = toc (id);
%!
%!  assert (d, max (sd, 0), sqrt (eps));
%!  printf ("ls_distance_fcn faster than ls_signed_distance by %.1f%%\n", ...
%!          100 * ((time2 - time1) / time2));
%!  assert (time1 < time2);

Since the timing is measured by tic/toc, there is no guarantee that other jobs will not disturb the test.

This is problematic for distributions like Debian that relies on rebuilding the package on different architectures.  Here is the build log for the octave-level-set on a Debian i386 system:
https://buildd.debian.org/status/fetch.php?pkg=octave-level-set&arch=i386&ver=0.3.0-10&stamp=1605086516&raw=0